/lib/python3.11/site-packages/torch/jit/_script.py:1491: FutureWarning: `torch.jit.script` is deprecated. Please switch to `torch.compile` or `torch.export`. warnings.warn( /lib/python3.11/site-packages/mujoco_warp/_src/io.py:713: UserWarning: MULTICCD is enabled, but the scene contains CCD pairs without multicontact support: [('CYLINDER', 'CYLINDER'), ('CYLINDER', 'BOX')]. At most 1 contact will be generated for these pairs. warnings.warn( Setting seed: 0 Warp 1.14.0 initialized: CUDA Toolkit 12.9, Driver 13.0 Devices: "cpu" : "x86_64" "cuda:0" : "NVIDIA GeForce RTX 5070 Ti" (15 GiB, sm_120, mempool enabled) Kernel cache: /1.14.0 Module mujoco_warp._src.smooth e597d07 load on device 'cuda:0' took 1.97 ms (cached) Module _nxn_broadphase__locals__kernel_df6640ed df6640e load on device 'cuda:0' took 1.01 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_f3c975b7 f3c975b load on device 'cuda:0' took 1.14 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_f3c975b7 fa59860 load on device 'cuda:0' took 0.98 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_82ff21e7 82ff21e load on device 'cuda:0' took 0.85 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_82ff21e7 fa618bd load on device 'cuda:0' took 0.95 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_240d51e8 240d51e load on device 'cuda:0' took 1.22 ms (cached) Module ccd_kernel_builder__locals__ccd_kernel_240d51e8 a5c9e1e load on device 'cuda:0' took 1.08 ms (cached) Module _primitive_narrowphase__locals__primitive_narrowphase_9aaa7d98 9aaa7d9 load on device 'cuda:0' took 0.44 ms (cached) Module mujoco_warp._src.constraint 880bd7f load on device 'cuda:0' took 0.39 ms (cached) Module _friction_dof__locals__kernel_3c14efda 3c14efd load on device 'cuda:0' took 0.28 ms (cached) Module _limit_slide_hinge__locals__kernel_4d6c55ab 4d6c55a load on device 'cuda:0' took 0.47 ms (cached) Module _efc_contact_init__locals__kernel_744a1fda 744a1fd load on device 'cuda:0' took 0.44 ms (cached) Module _efc_contact_jac_dense__locals__kernel_760e2e1a d6aba00 load on device 'cuda:0' took 0.45 ms (cached) Module _efc_contact_update__locals__kernel_c40b0799 c40b079 load on device 'cuda:0' took 0.44 ms (cached) Module mujoco_warp._src.forward 687ddb7 load on device 'cuda:0' took 0.80 ms (cached) Module mujoco_warp._src.passive 8ff6993 load on device 'cuda:0' took 0.89 ms (cached) Module mujoco_warp._src.forward 2190340 load on device 'cuda:0' took 1.03 ms (cached) Module _qfrc_smooth__locals__kernel_6eb3d958 6eb3d95 load on device 'cuda:0' took 0.34 ms (cached) Module mujoco_warp._src.support d1e71d9 load on device 'cuda:0' took 0.24 ms (cached) Module _tile_cholesky_factorize_solve_block__locals__kernel_b44bd18c 0966f64 load on device 'cuda:0' took 16.74 ms (cached) Module _solve_init_dof__locals__kernel_99a497d1 99a497d load on device 'cuda:0' took 0.27 ms (cached) Module mujoco_warp._src.solver 245b5b3 load on device 'cuda:0' took 1.12 ms (cached) Module _solve_init_jaref_kernel__locals__kernel_d620bcf9 d620bcf load on device 'cuda:0' took 0.37 ms (cached) Module mul_m_kernel__locals___mul_m_fd2186c4 fd2186c load on device 'cuda:0' took 0.41 ms (cached) Module _update_constraint_efc__locals__kernel_4af8f110 4af8f11 load on device 'cuda:0' took 0.33 ms (cached) Module _update_constraint_init_qfrc_constraint_dense__locals__kernel_391cb5b4 391cb5b load on device 'cuda:0' took 0.40 ms (cached) Module _update_gradient_zero_grad_dot__locals__kernel_cb780630 cb78063 load on device 'cuda:0' took 0.40 ms (cached) Module _update_gradient_grad__locals__kernel_0cbf4927 0cbf492 load on device 'cuda:0' took 0.30 ms (cached) Module _update_gradient_JTDAJ_dense_tiled__locals__kernel_5d6c5d76 c826079 load on device 'cuda:0' took 0.52 ms (cached) Module _update_gradient_cholesky__locals__kernel_b27fc0c0 f9b150c load on device 'cuda:0' took 17.63 ms (cached) Module _linesearch_iterative_kernel__locals__kernel_d3ea36af e1537a9 load on device 'cuda:0' took 0.59 ms (cached) Module mujoco_warp._src.sensor 42e0a24 load on device 'cuda:0' took 1.33 ms (cached) Module _contact_sort__locals__contact_sort_321f5f3e 33ea4e1 load on device 'cuda:0' took 0.51 ms (cached) Module mujoco_warp._src.derivative e770e28 load on device 'cuda:0' took 0.72 ms (cached) Module _next_time_builder__locals___next_time_95ad6d7a 95ad6d7 load on device 'cuda:0' took 0.36 ms (cached) Module reset_data__locals__reset_xfrc_applied_03030511 0303051 load on device 'cuda:0' took 0.44 ms (cached) Module reset_data__locals__reset_M_631075dd 631075d load on device 'cuda:0' took 0.28 ms (cached) Module reset_data__locals__reset_mocap_6513201c 6513201 load on device 'cuda:0' took 0.31 ms (cached) Module reset_data__locals__reset_contact_5c08f689 5c08f68 load on device 'cuda:0' took 0.41 ms (cached) Module reset_data__locals__reset_sleep_50b990da 50b990d load on device 'cuda:0' took 0.54 ms (cached) Module reset_data__locals__reset_nworld_69a10925 69a1092 load on device 'cuda:0' took 0.42 ms (cached) +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 4096 | | Environment device | cuda:0 | | Environment seed | 0 | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 2 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ +---------------------------------------+ | Active Event Terms in Mode: 'startup' | +-----------+---------------------------+ | Index | Name | +-----------+---------------------------+ | 0 | foot_friction | | 1 | encoder_bias | | 2 | base_com | +-----------+---------------------------+ +-----------------------------+ | Domain Randomization Fields | +-------+---------------------+ | Index | Field Name | +-------+---------------------+ | 0 | geom_friction | | 1 | body_ipos | | 2 | body_subtreemass | | 3 | dof_invweight0 | | 4 | body_invweight0 | | 5 | tendon_length0 | | 6 | tendon_invweight0 | | 7 | actuator_acc0 | +-------+---------------------+ Module repeat_array_kernel_4f67a4c3 711f54c load on device 'cuda:0' took 0.34 ms (cached) Module repeat_array_kernel_4f67a4c3 7224d07 load on device 'cuda:0' took 0.42 ms (cached) Module repeat_array_kernel_4f67a4c3 b2885b0 load on device 'cuda:0' took 0.44 ms (cached) [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 2 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ +------------------------------------------------------------+ | Active Observation Terms in Group: 'critic' (shape: (56,)) | +-----------+----------------------------------+-------------+ | Index | Name | Shape | +-----------+----------------------------------+-------------+ | 0 | base_ang_vel | (3,) | | 1 | projected_gravity | (3,) | | 2 | joint_pos | (12,) | | 3 | joint_vel | (12,) | | 4 | wheel_vel | (4,) | | 5 | actions | (16,) | | 6 | command | (3,) | | 7 | base_lin_vel | (3,) | +-----------+----------------------------------+-------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 14 active terms. +-----------------------------------------+ | Active Reward Terms | +-------+------------------------+--------+ | Index | Name | Weight | +-------+------------------------+--------+ | 0 | track_linear_velocity | 2.0 | | 1 | track_angular_velocity | 2.0 | | 2 | upright | 1.0 | | 3 | leg_posture | -0.4 | | 4 | standstill | -2.0 | | 5 | action_rate_l2 | -0.05 | | 6 | joint_torques_l2 | -2e-05 | | 7 | dof_pos_limits | -1.0 | | 8 | terminated | -20.0 | | 9 | unnecessary_wheel_lift | -0.2 | | 10 | rolling_leg_motion | -0.12 | | 11 | obstacle_progress | 0.5 | | 12 | wheel_body_intrusion | -1.0 | | 13 | pair_body_intrusion | -1.0 | +-------+------------------------+--------+ [INFO] contains 1 active terms. +-------------------------+ | Active Curriculum Terms | +-----------+-------------+ | Index | Name | +-----------+-------------+ | 0 | course | +-----------+-------------+ [INFO] contains 1 active terms. +-------------------------+ | Active Metrics Terms | +-------+-----------------+ | Index | Name | +-------+-----------------+ | 0 | mean_action_acc | +-------+-----------------+ [INFO] (inactive) Module mujoco_warp._src.io 19274fb load on device 'cuda:0' took 1.04 ms (cached) Module _tile_cholesky_factorize_block__locals__kernel_ccbdb7fb c86e457 load on device 'cuda:0' took 13.62 ms (cached) Module _tile_cholesky_solve_block__locals__kernel_320f9d39 639ca7f load on device 'cuda:0' took 13.33 ms (cached) Shared terrain atlas: 21.6 MiB S10 replay storage: 3424.0 MiB +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_000000.mp4 ============================================================================== Learning iteration 50/50000 Total steps: 204,800 Steps per second: 7,596 Iteration time: 0.539s Collection time: 0.104s Learning time: 0.307s Mean reward: +0.0671 Mean episode length: 46.3 Actor loss: n/a Critic loss: n/a World model loss: +3.3125 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +1.1947 Episode_Reward/track_angular_velocity: +1.5756 Episode_Reward/upright: +0.8977 Episode_Reward/leg_posture: -0.1223 Episode_Reward/standstill: -0.3052 Episode_Reward/action_rate_l2: -0.0017 Episode_Reward/joint_torques_l2: -0.0031 Episode_Reward/dof_pos_limits: -0.0000 Episode_Reward/terminated: -0.0010 Episode_Reward/unnecessary_wheel_lift: -0.0605 Episode_Reward/rolling_leg_motion: -0.0584 Episode_Reward/obstacle_progress: +0.0000 Episode_Reward/wheel_body_intrusion: -0.0108 Episode_Reward/pair_body_intrusion: -0.0019 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1936 Metrics/base_velocity/error_vel_yaw: 0.2417 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0909 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.9091 ============================================================================== ============================================================================== Learning iteration 100/50000 Total steps: 409,600 Steps per second: 8,140 Iteration time: 0.503s Collection time: 0.108s Learning time: 0.358s Mean reward: +0.0493 Mean episode length: 80.0 Actor loss: n/a Critic loss: n/a World model loss: +3.7498 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +1.8374 Episode_Reward/track_angular_velocity: +2.5339 Episode_Reward/upright: +1.3571 Episode_Reward/leg_posture: -0.5664 Episode_Reward/standstill: -0.7668 Episode_Reward/action_rate_l2: -0.0029 Episode_Reward/joint_torques_l2: -0.0109 Episode_Reward/dof_pos_limits: -0.0033 Episode_Reward/terminated: -0.1579 Episode_Reward/unnecessary_wheel_lift: -0.0535 Episode_Reward/rolling_leg_motion: -0.0484 Episode_Reward/obstacle_progress: -0.0000 Episode_Reward/wheel_body_intrusion: -0.1348 Episode_Reward/pair_body_intrusion: -0.0420 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2380 Metrics/base_velocity/error_vel_yaw: 0.3078 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.2990 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.8010 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1966] ============================================================================== Learning iteration 150/50000 Total steps: 614,400 Steps per second: 8,360 Iteration time: 0.490s Collection time: 0.109s Learning time: 0.355s Mean reward: +0.0493 Mean episode length: 119.8 Actor loss: n/a Critic loss: n/a World model loss: +3.0211 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +2.8006 Episode_Reward/track_angular_velocity: +3.8681 Episode_Reward/upright: +2.0143 Episode_Reward/leg_posture: -0.8256 Episode_Reward/standstill: -1.2031 Episode_Reward/action_rate_l2: -0.0043 Episode_Reward/joint_torques_l2: -0.0158 Episode_Reward/dof_pos_limits: -0.0089 Episode_Reward/terminated: -0.2957 Episode_Reward/unnecessary_wheel_lift: -0.1188 Episode_Reward/rolling_leg_motion: -0.0822 Episode_Reward/obstacle_progress: -0.0001 Episode_Reward/wheel_body_intrusion: -0.1721 Episode_Reward/pair_body_intrusion: -0.0482 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2314 Metrics/base_velocity/error_vel_yaw: 0.2903 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4106 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7674 ============================================================================== ============================================================================== Learning iteration 200/50000 Total steps: 819,200 Steps per second: 8,461 Iteration time: 0.484s Collection time: 0.110s Learning time: 0.356s Mean reward: +0.0500 Mean episode length: 139.9 Actor loss: n/a Critic loss: n/a World model loss: +2.5222 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +3.2475 Episode_Reward/track_angular_velocity: +4.5193 Episode_Reward/upright: +2.3730 Episode_Reward/leg_posture: -0.9561 Episode_Reward/standstill: -1.3552 Episode_Reward/action_rate_l2: -0.0051 Episode_Reward/joint_torques_l2: -0.0183 Episode_Reward/dof_pos_limits: -0.0095 Episode_Reward/terminated: -0.3176 Episode_Reward/unnecessary_wheel_lift: -0.1258 Episode_Reward/rolling_leg_motion: -0.0944 Episode_Reward/obstacle_progress: +0.0002 Episode_Reward/wheel_body_intrusion: -0.2021 Episode_Reward/pair_body_intrusion: -0.0529 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2339 Metrics/base_velocity/error_vel_yaw: 0.2904 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3916 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7814 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2386] ============================================================================== Learning iteration 250/50000 Total steps: 1,024,000 Steps per second: 8,568 Iteration time: 0.478s Collection time: 0.110s Learning time: 0.344s Mean reward: +0.0505 Mean episode length: 148.2 Actor loss: n/a Critic loss: n/a World model loss: +2.1572 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +3.4519 Episode_Reward/track_angular_velocity: +4.7714 Episode_Reward/upright: +2.5070 Episode_Reward/leg_posture: -1.0174 Episode_Reward/standstill: -1.3489 Episode_Reward/action_rate_l2: -0.0054 Episode_Reward/joint_torques_l2: -0.0194 Episode_Reward/dof_pos_limits: -0.0098 Episode_Reward/terminated: -0.3486 Episode_Reward/unnecessary_wheel_lift: -0.1346 Episode_Reward/rolling_leg_motion: -0.0987 Episode_Reward/obstacle_progress: -0.0000 Episode_Reward/wheel_body_intrusion: -0.2058 Episode_Reward/pair_body_intrusion: -0.0601 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2333 Metrics/base_velocity/error_vel_yaw: 0.2940 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4120 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7542 ============================================================================== ============================================================================== Learning iteration 300/50000 Total steps: 1,228,800 Steps per second: 8,608 Iteration time: 0.476s Collection time: 0.110s Learning time: 0.355s Mean reward: +0.0507 Mean episode length: 155.8 Actor loss: n/a Critic loss: n/a World model loss: +1.8894 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +3.6528 Episode_Reward/track_angular_velocity: +5.0241 Episode_Reward/upright: +2.6621 Episode_Reward/leg_posture: -1.0477 Episode_Reward/standstill: -1.4823 Episode_Reward/action_rate_l2: -0.0056 Episode_Reward/joint_torques_l2: -0.0201 Episode_Reward/dof_pos_limits: -0.0092 Episode_Reward/terminated: -0.3467 Episode_Reward/unnecessary_wheel_lift: -0.1335 Episode_Reward/rolling_leg_motion: -0.1051 Episode_Reward/obstacle_progress: +0.0002 Episode_Reward/wheel_body_intrusion: -0.2214 Episode_Reward/pair_body_intrusion: -0.0612 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2312 Metrics/base_velocity/error_vel_yaw: 0.2926 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3591 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.8051 ============================================================================== /lib/python3.11/site-packages/triton/knobs.py:504: UserWarning: TRITON_DEFAULT_FP_FUSION is unset: defaulting to FP fusion ENABLED for backward compatibility in Triton 3.9. The default will change to DISABLED in Triton 3.10; set TRITON_DEFAULT_FP_FUSION=1 to keep fusion enabled or =0 to adopt the new default early and silence this warning. warn_once(FP_FUSION_DEFAULT_WARNING) ============================================================================== Learning iteration 350/50000 Total steps: 1,433,600 Steps per second: 7,614 Iteration time: 0.538s Collection time: 0.110s Learning time: 0.800s Mean reward: +0.0507 Mean episode length: 160.9 Actor loss: -1.9025 Critic loss: +0.0733 World model loss: +1.7612 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +3.7773 Episode_Reward/track_angular_velocity: +5.2195 Episode_Reward/upright: +2.7272 Episode_Reward/leg_posture: -1.0968 Episode_Reward/standstill: -1.5107 Episode_Reward/action_rate_l2: -0.0058 Episode_Reward/joint_torques_l2: -0.0210 Episode_Reward/dof_pos_limits: -0.0111 Episode_Reward/terminated: -0.3742 Episode_Reward/unnecessary_wheel_lift: -0.1427 Episode_Reward/rolling_leg_motion: -0.1065 Episode_Reward/obstacle_progress: -0.0002 Episode_Reward/wheel_body_intrusion: -0.2237 Episode_Reward/pair_body_intrusion: -0.0676 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2308 Metrics/base_velocity/error_vel_yaw: 0.2862 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3720 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7851 ============================================================================== ============================================================================== Learning iteration 400/50000 Total steps: 1,638,400 Steps per second: 7,610 Iteration time: 0.538s Collection time: 0.110s Learning time: 0.430s Mean reward: +0.0531 Mean episode length: 169.3 Actor loss: -2.2357 Critic loss: +0.0583 World model loss: +1.5993 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +4.0490 Episode_Reward/track_angular_velocity: +5.5182 Episode_Reward/upright: +2.9128 Episode_Reward/leg_posture: -1.1052 Episode_Reward/standstill: -1.4661 Episode_Reward/action_rate_l2: -0.0061 Episode_Reward/joint_torques_l2: -0.0212 Episode_Reward/dof_pos_limits: -0.0098 Episode_Reward/terminated: -0.3455 Episode_Reward/unnecessary_wheel_lift: -0.1389 Episode_Reward/rolling_leg_motion: -0.1105 Episode_Reward/obstacle_progress: +0.0003 Episode_Reward/wheel_body_intrusion: -0.2217 Episode_Reward/pair_body_intrusion: -0.0662 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2250 Metrics/base_velocity/error_vel_yaw: 0.2821 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3722 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7990 ============================================================================== ============================================================================== Learning iteration 450/50000 Total steps: 1,843,200 Steps per second: 7,601 Iteration time: 0.539s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0545 Mean episode length: 173.3 Actor loss: -2.6894 Critic loss: +0.0542 World model loss: +1.4312 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +4.2831 Episode_Reward/track_angular_velocity: +5.7092 Episode_Reward/upright: +3.0359 Episode_Reward/leg_posture: -1.0592 Episode_Reward/standstill: -1.6052 Episode_Reward/action_rate_l2: -0.0063 Episode_Reward/joint_torques_l2: -0.0203 Episode_Reward/dof_pos_limits: -0.0087 Episode_Reward/terminated: -0.3341 Episode_Reward/unnecessary_wheel_lift: -0.1361 Episode_Reward/rolling_leg_motion: -0.1142 Episode_Reward/obstacle_progress: +0.0011 Episode_Reward/wheel_body_intrusion: -0.2350 Episode_Reward/pair_body_intrusion: -0.0737 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2161 Metrics/base_velocity/error_vel_yaw: 0.2723 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3313 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.8430 ============================================================================== ============================================================================== Learning iteration 500/50000 Total steps: 2,048,000 Steps per second: 7,596 Iteration time: 0.539s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0522 Mean episode length: 184.2 Actor loss: -3.2402 Critic loss: +0.0538 World model loss: +1.2875 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +4.4792 Episode_Reward/track_angular_velocity: +6.0436 Episode_Reward/upright: +3.3031 Episode_Reward/leg_posture: -1.0328 Episode_Reward/standstill: -2.2935 Episode_Reward/action_rate_l2: -0.0069 Episode_Reward/joint_torques_l2: -0.0203 Episode_Reward/dof_pos_limits: -0.0073 Episode_Reward/terminated: -0.2965 Episode_Reward/unnecessary_wheel_lift: -0.1323 Episode_Reward/rolling_leg_motion: -0.1271 Episode_Reward/obstacle_progress: +0.0105 Episode_Reward/wheel_body_intrusion: -0.2197 Episode_Reward/pair_body_intrusion: -0.0897 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2215 Metrics/base_velocity/error_vel_yaw: 0.2768 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.2682 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.8859 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2239] ============================================================================== Learning iteration 550/50000 Total steps: 2,252,800 Steps per second: 7,587 Iteration time: 0.540s Collection time: 0.111s Learning time: 0.434s Mean reward: +0.0554 Mean episode length: 190.8 Actor loss: -3.5373 Critic loss: +0.0417 World model loss: +1.2541 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +4.8436 Episode_Reward/track_angular_velocity: +6.2498 Episode_Reward/upright: +3.4885 Episode_Reward/leg_posture: -0.9548 Episode_Reward/standstill: -2.1755 Episode_Reward/action_rate_l2: -0.0074 Episode_Reward/joint_torques_l2: -0.0194 Episode_Reward/dof_pos_limits: -0.0062 Episode_Reward/terminated: -0.2743 Episode_Reward/unnecessary_wheel_lift: -0.1343 Episode_Reward/rolling_leg_motion: -0.1371 Episode_Reward/obstacle_progress: +0.0385 Episode_Reward/wheel_body_intrusion: -0.2233 Episode_Reward/pair_body_intrusion: -0.1215 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2087 Metrics/base_velocity/error_vel_yaw: 0.2791 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1957 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.9321 ============================================================================== ============================================================================== Learning iteration 600/50000 Total steps: 2,457,600 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0599 Mean episode length: 208.7 Actor loss: -4.1602 Critic loss: +0.0695 World model loss: +1.2228 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +5.5726 Episode_Reward/track_angular_velocity: +6.8743 Episode_Reward/upright: +3.8732 Episode_Reward/leg_posture: -0.9660 Episode_Reward/standstill: -2.1803 Episode_Reward/action_rate_l2: -0.0087 Episode_Reward/joint_torques_l2: -0.0199 Episode_Reward/dof_pos_limits: -0.0041 Episode_Reward/terminated: -0.1659 Episode_Reward/unnecessary_wheel_lift: -0.0894 Episode_Reward/rolling_leg_motion: -0.1583 Episode_Reward/obstacle_progress: +0.0661 Episode_Reward/wheel_body_intrusion: -0.1952 Episode_Reward/pair_body_intrusion: -0.0869 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1953 Metrics/base_velocity/error_vel_yaw: 0.2740 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.2138 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.9312 ============================================================================== ============================================================================== Learning iteration 650/50000 Total steps: 2,662,400 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.110s Learning time: 0.430s Mean reward: +0.0562 Mean episode length: 223.2 Actor loss: -4.6352 Critic loss: +0.0523 World model loss: +1.1959 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +5.8220 Episode_Reward/track_angular_velocity: +7.2501 Episode_Reward/upright: +4.1404 Episode_Reward/leg_posture: -1.0298 Episode_Reward/standstill: -2.9596 Episode_Reward/action_rate_l2: -0.0109 Episode_Reward/joint_torques_l2: -0.0214 Episode_Reward/dof_pos_limits: -0.0038 Episode_Reward/terminated: -0.1395 Episode_Reward/unnecessary_wheel_lift: -0.1190 Episode_Reward/rolling_leg_motion: -0.2150 Episode_Reward/obstacle_progress: +0.0818 Episode_Reward/wheel_body_intrusion: -0.1692 Episode_Reward/pair_body_intrusion: -0.0737 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.2042 Metrics/base_velocity/error_vel_yaw: 0.2858 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4531 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.8219 ============================================================================== ============================================================================== Learning iteration 700/50000 Total steps: 2,867,200 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.110s Learning time: 0.431s Mean reward: +0.0570 Mean episode length: 237.5 Actor loss: -5.1414 Critic loss: +0.0611 World model loss: +1.1838 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +6.4777 Episode_Reward/track_angular_velocity: +7.7589 Episode_Reward/upright: +4.4663 Episode_Reward/leg_posture: -0.9403 Episode_Reward/standstill: -3.5809 Episode_Reward/action_rate_l2: -0.0159 Episode_Reward/joint_torques_l2: -0.0203 Episode_Reward/dof_pos_limits: -0.0040 Episode_Reward/terminated: -0.1280 Episode_Reward/unnecessary_wheel_lift: -0.1213 Episode_Reward/rolling_leg_motion: -0.3152 Episode_Reward/obstacle_progress: +0.1165 Episode_Reward/wheel_body_intrusion: -0.1046 Episode_Reward/pair_body_intrusion: -0.0509 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1905 Metrics/base_velocity/error_vel_yaw: 0.2804 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.5725 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7319 ============================================================================== ============================================================================== Learning iteration 750/50000 Total steps: 3,072,000 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.110s Learning time: 0.430s Mean reward: +0.0622 Mean episode length: 285.7 Actor loss: -5.5572 Critic loss: +0.0378 World model loss: +1.1114 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +8.1537 Episode_Reward/track_angular_velocity: +9.5306 Episode_Reward/upright: +5.4430 Episode_Reward/leg_posture: -1.0268 Episode_Reward/standstill: -3.6588 Episode_Reward/action_rate_l2: -0.0199 Episode_Reward/joint_torques_l2: -0.0217 Episode_Reward/dof_pos_limits: -0.0043 Episode_Reward/terminated: -0.1161 Episode_Reward/unnecessary_wheel_lift: -0.1400 Episode_Reward/rolling_leg_motion: -0.3917 Episode_Reward/obstacle_progress: +0.1675 Episode_Reward/wheel_body_intrusion: -0.0823 Episode_Reward/pair_body_intrusion: -0.0582 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1773 Metrics/base_velocity/error_vel_yaw: 0.2627 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.6106 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.6154 ============================================================================== ============================================================================== Learning iteration 800/50000 Total steps: 3,276,800 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.110s Learning time: 0.431s Mean reward: +0.0657 Mean episode length: 300.9 Actor loss: -5.9100 Critic loss: +0.0253 World model loss: +1.0840 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +8.8683 Episode_Reward/track_angular_velocity: +10.1152 Episode_Reward/upright: +5.7353 Episode_Reward/leg_posture: -1.2377 Episode_Reward/standstill: -3.0438 Episode_Reward/action_rate_l2: -0.0202 Episode_Reward/joint_torques_l2: -0.0232 Episode_Reward/dof_pos_limits: -0.0050 Episode_Reward/terminated: -0.0840 Episode_Reward/unnecessary_wheel_lift: -0.1268 Episode_Reward/rolling_leg_motion: -0.3739 Episode_Reward/obstacle_progress: +0.1872 Episode_Reward/wheel_body_intrusion: -0.1207 Episode_Reward/pair_body_intrusion: -0.1043 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1658 Metrics/base_velocity/error_vel_yaw: 0.2561 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4965 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7762 ============================================================================== ============================================================================== Learning iteration 850/50000 Total steps: 3,481,600 Steps per second: 7,581 Iteration time: 0.540s Collection time: 0.109s Learning time: 0.432s Mean reward: +0.0662 Mean episode length: 345.1 Actor loss: -6.2149 Critic loss: +0.0234 World model loss: +1.0936 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +10.3925 Episode_Reward/track_angular_velocity: +11.6744 Episode_Reward/upright: +6.5685 Episode_Reward/leg_posture: -1.5710 Episode_Reward/standstill: -3.3495 Episode_Reward/action_rate_l2: -0.0262 Episode_Reward/joint_torques_l2: -0.0283 Episode_Reward/dof_pos_limits: -0.0042 Episode_Reward/terminated: -0.0964 Episode_Reward/unnecessary_wheel_lift: -0.1419 Episode_Reward/rolling_leg_motion: -0.5353 Episode_Reward/obstacle_progress: +0.2366 Episode_Reward/wheel_body_intrusion: -0.1567 Episode_Reward/pair_body_intrusion: -0.1181 Curriculum/terrain_level_mean: 0.000e+00 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1592 Metrics/base_velocity/error_vel_yaw: 0.2511 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3636 Episode_Termination/course_exit: 0.0000 Episode_Termination/illegal_nonwheel_contact: 0.7552 ============================================================================== ============================================================================== Learning iteration 900/50000 Total steps: 3,686,400 Steps per second: 7,582 Iteration time: 0.540s Collection time: 0.109s Learning time: 0.430s Mean reward: +0.0716 Mean episode length: 392.6 Actor loss: -6.6069 Critic loss: +0.0227 World model loss: +1.0809 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +12.4170 Episode_Reward/track_angular_velocity: +13.6012 Episode_Reward/upright: +7.5330 Episode_Reward/leg_posture: -1.6680 Episode_Reward/standstill: -3.0011 Episode_Reward/action_rate_l2: -0.0323 Episode_Reward/joint_torques_l2: -0.0337 Episode_Reward/dof_pos_limits: -0.0045 Episode_Reward/terminated: -0.0590 Episode_Reward/unnecessary_wheel_lift: -0.1062 Episode_Reward/rolling_leg_motion: -0.6175 Episode_Reward/obstacle_progress: +0.2904 Episode_Reward/wheel_body_intrusion: -0.1745 Episode_Reward/pair_body_intrusion: -0.0239 Curriculum/terrain_level_mean: 4.883e-04 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1421 Metrics/base_velocity/error_vel_yaw: 0.2288 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3544 Episode_Termination/course_exit: 0.0253 Episode_Termination/illegal_nonwheel_contact: 0.7722 ============================================================================== ============================================================================== Learning iteration 950/50000 Total steps: 3,891,200 Steps per second: 7,580 Iteration time: 0.540s Collection time: 0.109s Learning time: 0.433s Mean reward: +0.0758 Mean episode length: 453.3 Actor loss: -6.9023 Critic loss: +0.0163 World model loss: +1.0748 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +14.9481 Episode_Reward/track_angular_velocity: +16.0827 Episode_Reward/upright: +8.7521 Episode_Reward/leg_posture: -1.8947 Episode_Reward/standstill: -2.7401 Episode_Reward/action_rate_l2: -0.0380 Episode_Reward/joint_torques_l2: -0.0401 Episode_Reward/dof_pos_limits: -0.0059 Episode_Reward/terminated: -0.0513 Episode_Reward/unnecessary_wheel_lift: -0.0923 Episode_Reward/rolling_leg_motion: -0.6802 Episode_Reward/obstacle_progress: +0.3504 Episode_Reward/wheel_body_intrusion: -0.1886 Episode_Reward/pair_body_intrusion: -0.0206 Curriculum/terrain_level_mean: +0.0015 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1267 Metrics/base_velocity/error_vel_yaw: 0.2059 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4839 Episode_Termination/course_exit: 0.0645 Episode_Termination/illegal_nonwheel_contact: 0.6935 ============================================================================== ============================================================================== Learning iteration 1000/50000 Total steps: 4,096,000 Steps per second: 7,575 Iteration time: 0.541s Collection time: 0.108s Learning time: 0.439s Mean reward: +0.0773 Mean episode length: 494.1 Actor loss: -7.0796 Critic loss: +0.0123 World model loss: +1.0596 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +16.6281 Episode_Reward/track_angular_velocity: +17.6545 Episode_Reward/upright: +9.5628 Episode_Reward/leg_posture: -2.2200 Episode_Reward/standstill: -2.7688 Episode_Reward/action_rate_l2: -0.0369 Episode_Reward/joint_torques_l2: -0.0438 Episode_Reward/dof_pos_limits: -0.0064 Episode_Reward/terminated: -0.0396 Episode_Reward/unnecessary_wheel_lift: -0.1022 Episode_Reward/rolling_leg_motion: -0.4940 Episode_Reward/obstacle_progress: +0.3648 Episode_Reward/wheel_body_intrusion: -0.2402 Episode_Reward/pair_body_intrusion: -0.0412 Curriculum/terrain_level_mean: +0.0039 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1191 Metrics/base_velocity/error_vel_yaw: 0.1992 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.4314 Episode_Termination/course_exit: 0.1961 Episode_Termination/illegal_nonwheel_contact: 0.5294 ============================================================================== ============================================================================== Learning iteration 1050/50000 Total steps: 4,300,800 Steps per second: 7,574 Iteration time: 0.541s Collection time: 0.108s Learning time: 0.435s Mean reward: +0.0762 Mean episode length: 565.7 Actor loss: -7.3239 Critic loss: +0.0114 World model loss: +1.0179 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +19.0366 Episode_Reward/track_angular_velocity: +20.2239 Episode_Reward/upright: +10.8895 Episode_Reward/leg_posture: -3.0924 Episode_Reward/standstill: -3.1703 Episode_Reward/action_rate_l2: -0.0396 Episode_Reward/joint_torques_l2: -0.0545 Episode_Reward/dof_pos_limits: -0.0079 Episode_Reward/terminated: -0.0431 Episode_Reward/unnecessary_wheel_lift: -0.1027 Episode_Reward/rolling_leg_motion: -0.4314 Episode_Reward/obstacle_progress: +0.3649 Episode_Reward/wheel_body_intrusion: -0.3643 Episode_Reward/pair_body_intrusion: -0.1202 Curriculum/terrain_level_mean: +0.0098 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1187 Metrics/base_velocity/error_vel_yaw: 0.1983 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3594 Episode_Termination/course_exit: 0.3906 Episode_Termination/illegal_nonwheel_contact: 0.3281 ============================================================================== ============================================================================== Learning iteration 1100/50000 Total steps: 4,505,600 Steps per second: 7,572 Iteration time: 0.541s Collection time: 0.109s Learning time: 0.434s Mean reward: +0.0772 Mean episode length: 593.0 Actor loss: -7.4351 Critic loss: +0.0118 World model loss: +1.0231 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +20.1976 Episode_Reward/track_angular_velocity: +21.4841 Episode_Reward/upright: +11.4223 Episode_Reward/leg_posture: -3.2362 Episode_Reward/standstill: -3.3757 Episode_Reward/action_rate_l2: -0.0442 Episode_Reward/joint_torques_l2: -0.0591 Episode_Reward/dof_pos_limits: -0.0105 Episode_Reward/terminated: -0.0452 Episode_Reward/unnecessary_wheel_lift: -0.1009 Episode_Reward/rolling_leg_motion: -0.4222 Episode_Reward/obstacle_progress: +0.3599 Episode_Reward/wheel_body_intrusion: -0.3342 Episode_Reward/pair_body_intrusion: -0.0724 Curriculum/terrain_level_mean: +0.0156 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1138 Metrics/base_velocity/error_vel_yaw: 0.1850 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.3538 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.4154 ============================================================================== ============================================================================== Learning iteration 1150/50000 Total steps: 4,710,400 Steps per second: 7,572 Iteration time: 0.541s Collection time: 0.109s Learning time: 0.433s Mean reward: +0.0783 Mean episode length: 675.8 Actor loss: -7.6606 Critic loss: +0.0067 World model loss: +1.0190 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +23.2814 Episode_Reward/track_angular_velocity: +24.6759 Episode_Reward/upright: +13.0380 Episode_Reward/leg_posture: -3.4269 Episode_Reward/standstill: -3.8727 Episode_Reward/action_rate_l2: -0.0662 Episode_Reward/joint_torques_l2: -0.0674 Episode_Reward/dof_pos_limits: -0.0181 Episode_Reward/terminated: -0.0370 Episode_Reward/unnecessary_wheel_lift: -0.1031 Episode_Reward/rolling_leg_motion: -0.5566 Episode_Reward/obstacle_progress: +0.4161 Episode_Reward/wheel_body_intrusion: -0.2868 Episode_Reward/pair_body_intrusion: -0.0420 Curriculum/terrain_level_mean: +0.0271 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1099 Metrics/base_velocity/error_vel_yaw: 0.1769 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1500 Episode_Termination/course_exit: 0.6500 Episode_Termination/illegal_nonwheel_contact: 0.2375 ============================================================================== ============================================================================== Learning iteration 1200/50000 Total steps: 4,915,200 Steps per second: 7,569 Iteration time: 0.541s Collection time: 0.110s Learning time: 0.436s Mean reward: +0.0792 Mean episode length: 700.4 Actor loss: -7.7869 Critic loss: +0.0076 World model loss: +0.9589 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +24.3137 Episode_Reward/track_angular_velocity: +25.7290 Episode_Reward/upright: +13.5084 Episode_Reward/leg_posture: -3.3916 Episode_Reward/standstill: -3.9057 Episode_Reward/action_rate_l2: -0.0662 Episode_Reward/joint_torques_l2: -0.0701 Episode_Reward/dof_pos_limits: -0.0215 Episode_Reward/terminated: -0.0424 Episode_Reward/unnecessary_wheel_lift: -0.1448 Episode_Reward/rolling_leg_motion: -0.5107 Episode_Reward/obstacle_progress: +0.4095 Episode_Reward/wheel_body_intrusion: -0.2761 Episode_Reward/pair_body_intrusion: -0.0401 Curriculum/terrain_level_mean: +0.0391 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1067 Metrics/base_velocity/error_vel_yaw: 0.1698 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1928 Episode_Termination/course_exit: 0.6265 Episode_Termination/illegal_nonwheel_contact: 0.2169 ============================================================================== ============================================================================== Learning iteration 1250/50000 Total steps: 5,120,000 Steps per second: 7,566 Iteration time: 0.541s Collection time: 0.110s Learning time: 0.438s Mean reward: +0.0802 Mean episode length: 762.8 Actor loss: -7.9281 Critic loss: +0.0067 World model loss: +0.9403 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +26.7116 Episode_Reward/track_angular_velocity: +28.2094 Episode_Reward/upright: +14.7344 Episode_Reward/leg_posture: -3.6813 Episode_Reward/standstill: -3.9842 Episode_Reward/action_rate_l2: -0.0616 Episode_Reward/joint_torques_l2: -0.0773 Episode_Reward/dof_pos_limits: -0.0198 Episode_Reward/terminated: -0.0343 Episode_Reward/unnecessary_wheel_lift: -0.1236 Episode_Reward/rolling_leg_motion: -0.4916 Episode_Reward/obstacle_progress: +0.4218 Episode_Reward/wheel_body_intrusion: -0.3088 Episode_Reward/pair_body_intrusion: -0.0841 Curriculum/terrain_level_mean: +0.0522 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1622 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1519 Episode_Termination/course_exit: 0.7089 Episode_Termination/illegal_nonwheel_contact: 0.1519 ============================================================================== ============================================================================== Learning iteration 1300/50000 Total steps: 5,324,800 Steps per second: 7,563 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.437s Mean reward: +0.0795 Mean episode length: 783.3 Actor loss: -8.0810 Critic loss: +0.0048 World model loss: +0.8892 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +27.3859 Episode_Reward/track_angular_velocity: +28.9032 Episode_Reward/upright: +15.1066 Episode_Reward/leg_posture: -3.9549 Episode_Reward/standstill: -4.1074 Episode_Reward/action_rate_l2: -0.0589 Episode_Reward/joint_torques_l2: -0.0835 Episode_Reward/dof_pos_limits: -0.0210 Episode_Reward/terminated: -0.0459 Episode_Reward/unnecessary_wheel_lift: -0.1582 Episode_Reward/rolling_leg_motion: -0.5067 Episode_Reward/obstacle_progress: +0.4149 Episode_Reward/wheel_body_intrusion: -0.3965 Episode_Reward/pair_body_intrusion: -0.1751 Curriculum/terrain_level_mean: +0.0740 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1647 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1732 Episode_Termination/course_exit: 0.7638 Episode_Termination/illegal_nonwheel_contact: 0.1339 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1450] ============================================================================== Learning iteration 1350/50000 Total steps: 5,529,600 Steps per second: 7,561 Iteration time: 0.542s Collection time: 0.111s Learning time: 0.433s Mean reward: +0.0791 Mean episode length: 787.5 Actor loss: -8.1659 Critic loss: +0.0053 World model loss: +0.9595 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +27.4011 Episode_Reward/track_angular_velocity: +29.0405 Episode_Reward/upright: +15.1841 Episode_Reward/leg_posture: -3.9480 Episode_Reward/standstill: -4.1506 Episode_Reward/action_rate_l2: -0.0581 Episode_Reward/joint_torques_l2: -0.0895 Episode_Reward/dof_pos_limits: -0.0227 Episode_Reward/terminated: -0.0554 Episode_Reward/unnecessary_wheel_lift: -0.2045 Episode_Reward/rolling_leg_motion: -0.4965 Episode_Reward/obstacle_progress: +0.3883 Episode_Reward/wheel_body_intrusion: -0.4051 Episode_Reward/pair_body_intrusion: -0.2680 Curriculum/terrain_level_mean: +0.0901 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1650 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.1308 Episode_Termination/course_exit: 0.6636 Episode_Termination/illegal_nonwheel_contact: 0.2523 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1710] ============================================================================== Learning iteration 1400/50000 Total steps: 5,734,400 Steps per second: 7,560 Iteration time: 0.542s Collection time: 0.111s Learning time: 0.434s Mean reward: +0.0795 Mean episode length: 841.8 Actor loss: -8.2465 Critic loss: +0.0053 World model loss: +0.9002 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +29.3482 Episode_Reward/track_angular_velocity: +31.1705 Episode_Reward/upright: +16.2487 Episode_Reward/leg_posture: -4.0225 Episode_Reward/standstill: -4.4844 Episode_Reward/action_rate_l2: -0.0633 Episode_Reward/joint_torques_l2: -0.0966 Episode_Reward/dof_pos_limits: -0.0227 Episode_Reward/terminated: -0.0674 Episode_Reward/unnecessary_wheel_lift: -0.2486 Episode_Reward/rolling_leg_motion: -0.5127 Episode_Reward/obstacle_progress: +0.4200 Episode_Reward/wheel_body_intrusion: -0.4237 Episode_Reward/pair_body_intrusion: -0.3337 Curriculum/terrain_level_mean: +0.1084 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1035 Metrics/base_velocity/error_vel_yaw: 0.1602 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0977 Episode_Termination/course_exit: 0.6917 Episode_Termination/illegal_nonwheel_contact: 0.2256 ============================================================================== ============================================================================== Learning iteration 1450/50000 Total steps: 5,939,200 Steps per second: 7,559 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0802 Mean episode length: 835.6 Actor loss: -8.3086 Critic loss: +0.0041 World model loss: +0.8788 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +29.3016 Episode_Reward/track_angular_velocity: +30.9956 Episode_Reward/upright: +16.1671 Episode_Reward/leg_posture: -3.9575 Episode_Reward/standstill: -4.4286 Episode_Reward/action_rate_l2: -0.0622 Episode_Reward/joint_torques_l2: -0.0933 Episode_Reward/dof_pos_limits: -0.0183 Episode_Reward/terminated: -0.0326 Episode_Reward/unnecessary_wheel_lift: -0.1598 Episode_Reward/rolling_leg_motion: -0.4960 Episode_Reward/obstacle_progress: +0.4401 Episode_Reward/wheel_body_intrusion: -0.3791 Episode_Reward/pair_body_intrusion: -0.2968 Curriculum/terrain_level_mean: +0.1265 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.1007 Metrics/base_velocity/error_vel_yaw: 0.1582 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0556 Episode_Termination/course_exit: 0.8148 Episode_Termination/illegal_nonwheel_contact: 0.1574 ============================================================================== ============================================================================== Learning iteration 1500/50000 Total steps: 6,144,000 Steps per second: 7,558 Iteration time: 0.542s Collection time: 0.111s Learning time: 0.433s Mean reward: +0.0810 Mean episode length: 936.0 Actor loss: -8.3823 Critic loss: +0.0033 World model loss: +0.8955 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +32.9001 Episode_Reward/track_angular_velocity: +34.7749 Episode_Reward/upright: +18.1649 Episode_Reward/leg_posture: -4.1858 Episode_Reward/standstill: -4.7753 Episode_Reward/action_rate_l2: -0.0689 Episode_Reward/joint_torques_l2: -0.0998 Episode_Reward/dof_pos_limits: -0.0148 Episode_Reward/terminated: -0.0311 Episode_Reward/unnecessary_wheel_lift: -0.1553 Episode_Reward/rolling_leg_motion: -0.5592 Episode_Reward/obstacle_progress: +0.5039 Episode_Reward/wheel_body_intrusion: -0.3505 Episode_Reward/pair_body_intrusion: -0.2468 Curriculum/terrain_level_mean: +0.1414 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0999 Metrics/base_velocity/error_vel_yaw: 0.1562 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0805 Episode_Termination/course_exit: 0.8046 Episode_Termination/illegal_nonwheel_contact: 0.1379 ============================================================================== ============================================================================== Learning iteration 1550/50000 Total steps: 6,348,800 Steps per second: 7,554 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.440s Mean reward: +0.0813 Mean episode length: 982.3 Actor loss: -8.4340 Critic loss: +0.0031 World model loss: +0.8885 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +34.5657 Episode_Reward/track_angular_velocity: +36.5414 Episode_Reward/upright: +19.0772 Episode_Reward/leg_posture: -4.1793 Episode_Reward/standstill: -5.1171 Episode_Reward/action_rate_l2: -0.0750 Episode_Reward/joint_torques_l2: -0.1034 Episode_Reward/dof_pos_limits: -0.0138 Episode_Reward/terminated: -0.0345 Episode_Reward/unnecessary_wheel_lift: -0.2059 Episode_Reward/rolling_leg_motion: -0.6026 Episode_Reward/obstacle_progress: +0.5386 Episode_Reward/wheel_body_intrusion: -0.3248 Episode_Reward/pair_body_intrusion: -0.2005 Curriculum/terrain_level_mean: +0.1562 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0996 Metrics/base_velocity/error_vel_yaw: 0.1549 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0729 Episode_Termination/course_exit: 0.8125 Episode_Termination/illegal_nonwheel_contact: 0.1146 ============================================================================== ============================================================================== Learning iteration 1600/50000 Total steps: 6,553,600 Steps per second: 7,553 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0820 Mean episode length: 1003.7 Actor loss: -8.4996 Critic loss: +0.0033 World model loss: +0.9072 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +35.6237 Episode_Reward/track_angular_velocity: +37.4385 Episode_Reward/upright: +19.5210 Episode_Reward/leg_posture: -4.1029 Episode_Reward/standstill: -5.2297 Episode_Reward/action_rate_l2: -0.0795 Episode_Reward/joint_torques_l2: -0.1030 Episode_Reward/dof_pos_limits: -0.0104 Episode_Reward/terminated: -0.0274 Episode_Reward/unnecessary_wheel_lift: -0.1247 Episode_Reward/rolling_leg_motion: -0.6162 Episode_Reward/obstacle_progress: +0.5665 Episode_Reward/wheel_body_intrusion: -0.3199 Episode_Reward/pair_body_intrusion: -0.2114 Curriculum/terrain_level_mean: +0.1726 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0956 Metrics/base_velocity/error_vel_yaw: 0.1519 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.8526 Episode_Termination/illegal_nonwheel_contact: 0.1263 ============================================================================== ============================================================================== Learning iteration 1650/50000 Total steps: 6,758,400 Steps per second: 7,552 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.435s Mean reward: +0.0819 Mean episode length: 1089.1 Actor loss: -8.5813 Critic loss: +0.0024 World model loss: +0.8903 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +38.7161 Episode_Reward/track_angular_velocity: +40.7312 Episode_Reward/upright: +21.2240 Episode_Reward/leg_posture: -4.4238 Episode_Reward/standstill: -5.9804 Episode_Reward/action_rate_l2: -0.0889 Episode_Reward/joint_torques_l2: -0.1095 Episode_Reward/dof_pos_limits: -0.0047 Episode_Reward/terminated: -0.0149 Episode_Reward/unnecessary_wheel_lift: -0.1006 Episode_Reward/rolling_leg_motion: -0.6739 Episode_Reward/obstacle_progress: +0.6122 Episode_Reward/wheel_body_intrusion: -0.3757 Episode_Reward/pair_body_intrusion: -0.2669 Curriculum/terrain_level_mean: +0.1909 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1487 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0225 Episode_Termination/course_exit: 0.9213 Episode_Termination/illegal_nonwheel_contact: 0.0562 ============================================================================== ============================================================================== Learning iteration 1700/50000 Total steps: 6,963,200 Steps per second: 7,552 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0826 Mean episode length: 1074.1 Actor loss: -8.6507 Critic loss: +0.0025 World model loss: +0.8592 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +38.3826 Episode_Reward/track_angular_velocity: +40.2218 Episode_Reward/upright: +20.9475 Episode_Reward/leg_posture: -4.2699 Episode_Reward/standstill: -5.5707 Episode_Reward/action_rate_l2: -0.0889 Episode_Reward/joint_torques_l2: -0.1086 Episode_Reward/dof_pos_limits: -0.0025 Episode_Reward/terminated: -0.0147 Episode_Reward/unnecessary_wheel_lift: -0.1300 Episode_Reward/rolling_leg_motion: -0.6345 Episode_Reward/obstacle_progress: +0.6593 Episode_Reward/wheel_body_intrusion: -0.4530 Episode_Reward/pair_body_intrusion: -0.2645 Curriculum/terrain_level_mean: +0.2097 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1469 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.9263 Episode_Termination/illegal_nonwheel_contact: 0.0632 ============================================================================== ============================================================================== Learning iteration 1750/50000 Total steps: 7,168,000 Steps per second: 7,551 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.436s Mean reward: +0.0826 Mean episode length: 1400.0 Actor loss: -8.7178 Critic loss: +0.0018 World model loss: +0.9145 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9828 Episode_Reward/track_angular_velocity: +52.5681 Episode_Reward/upright: +27.3143 Episode_Reward/leg_posture: -5.6674 Episode_Reward/standstill: -7.3191 Episode_Reward/action_rate_l2: -0.1180 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0028 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.1422 Episode_Reward/rolling_leg_motion: -0.8202 Episode_Reward/obstacle_progress: +0.9026 Episode_Reward/wheel_body_intrusion: -0.6397 Episode_Reward/pair_body_intrusion: -0.3137 Curriculum/terrain_level_mean: +0.2288 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1441 Episode_Termination/time_out: 0.3960 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5705 Episode_Termination/illegal_nonwheel_contact: 0.0336 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[761] ============================================================================== Learning iteration 1800/50000 Total steps: 7,372,800 Steps per second: 7,550 Iteration time: 0.542s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0827 Mean episode length: 1193.0 Actor loss: -8.7484 Critic loss: +0.0020 World model loss: +0.8744 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.6801 Episode_Reward/track_angular_velocity: +44.8787 Episode_Reward/upright: +23.2946 Episode_Reward/leg_posture: -4.7039 Episode_Reward/standstill: -6.4700 Episode_Reward/action_rate_l2: -0.1025 Episode_Reward/joint_torques_l2: -0.1268 Episode_Reward/dof_pos_limits: -0.0032 Episode_Reward/terminated: -0.0047 Episode_Reward/unnecessary_wheel_lift: -0.1291 Episode_Reward/rolling_leg_motion: -0.7031 Episode_Reward/obstacle_progress: +0.7807 Episode_Reward/wheel_body_intrusion: -0.4868 Episode_Reward/pair_body_intrusion: -0.2630 Curriculum/terrain_level_mean: +0.2449 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1420 Episode_Termination/time_out: 0.0000 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.9767 Episode_Termination/illegal_nonwheel_contact: 0.0233 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2141] ============================================================================== Learning iteration 1850/50000 Total steps: 7,577,600 Steps per second: 7,549 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.435s Mean reward: +0.0825 Mean episode length: 1257.9 Actor loss: -8.8010 Critic loss: +0.0020 World model loss: +0.8594 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.0079 Episode_Reward/track_angular_velocity: +47.2924 Episode_Reward/upright: +24.5501 Episode_Reward/leg_posture: -5.0432 Episode_Reward/standstill: -6.9198 Episode_Reward/action_rate_l2: -0.1082 Episode_Reward/joint_torques_l2: -0.1362 Episode_Reward/dof_pos_limits: -0.0080 Episode_Reward/terminated: -0.0147 Episode_Reward/unnecessary_wheel_lift: -0.1542 Episode_Reward/rolling_leg_motion: -0.7278 Episode_Reward/obstacle_progress: +0.8475 Episode_Reward/wheel_body_intrusion: -0.5088 Episode_Reward/pair_body_intrusion: -0.2795 Curriculum/terrain_level_mean: +0.2629 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1422 Episode_Termination/time_out: 0.1495 Episode_Termination/fell_over: 0.0093 Episode_Termination/course_exit: 0.7944 Episode_Termination/illegal_nonwheel_contact: 0.0561 ============================================================================== ============================================================================== Learning iteration 1900/50000 Total steps: 7,782,400 Steps per second: 7,547 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.437s Mean reward: +0.0827 Mean episode length: 1319.2 Actor loss: -8.8392 Critic loss: +0.0014 World model loss: +0.8567 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0869 Episode_Reward/track_angular_velocity: +49.7337 Episode_Reward/upright: +25.7570 Episode_Reward/leg_posture: -5.1280 Episode_Reward/standstill: -7.2209 Episode_Reward/action_rate_l2: -0.1182 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0110 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.1798 Episode_Reward/rolling_leg_motion: -0.7789 Episode_Reward/obstacle_progress: +0.8788 Episode_Reward/wheel_body_intrusion: -0.4812 Episode_Reward/pair_body_intrusion: -0.2688 Curriculum/terrain_level_mean: +0.2876 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1390 Episode_Termination/time_out: 0.2403 Episode_Termination/fell_over: 0.0260 Episode_Termination/course_exit: 0.7078 Episode_Termination/illegal_nonwheel_contact: 0.0260 ============================================================================== ============================================================================== Learning iteration 1950/50000 Total steps: 7,987,200 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.437s Mean reward: +0.0833 Mean episode length: 1402.0 Actor loss: -8.9066 Critic loss: +0.0012 World model loss: +0.9010 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9937 Episode_Reward/track_angular_velocity: +53.0456 Episode_Reward/upright: +27.4139 Episode_Reward/leg_posture: -5.0238 Episode_Reward/standstill: -7.6162 Episode_Reward/action_rate_l2: -0.1291 Episode_Reward/joint_torques_l2: -0.1514 Episode_Reward/dof_pos_limits: -0.0085 Episode_Reward/terminated: -0.0164 Episode_Reward/unnecessary_wheel_lift: -0.1822 Episode_Reward/rolling_leg_motion: -0.8462 Episode_Reward/obstacle_progress: +0.9247 Episode_Reward/wheel_body_intrusion: -0.4308 Episode_Reward/pair_body_intrusion: -0.2533 Curriculum/terrain_level_mean: +0.3071 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.3262 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.6312 Episode_Termination/illegal_nonwheel_contact: 0.0213 ============================================================================== ============================================================================== Learning iteration 2000/50000 Total steps: 8,192,000 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.433s Mean reward: +0.0830 Mean episode length: 1468.5 Actor loss: -8.9329 Critic loss: +0.0015 World model loss: +0.8494 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +52.1304 Episode_Reward/track_angular_velocity: +55.6474 Episode_Reward/upright: +28.6664 Episode_Reward/leg_posture: -5.1921 Episode_Reward/standstill: -8.0848 Episode_Reward/action_rate_l2: -0.1411 Episode_Reward/joint_torques_l2: -0.1561 Episode_Reward/dof_pos_limits: -0.0231 Episode_Reward/terminated: -0.0344 Episode_Reward/unnecessary_wheel_lift: -0.2998 Episode_Reward/rolling_leg_motion: -0.8872 Episode_Reward/obstacle_progress: +1.0192 Episode_Reward/wheel_body_intrusion: -0.4324 Episode_Reward/pair_body_intrusion: -0.2601 Curriculum/terrain_level_mean: +0.3281 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4012 Episode_Termination/fell_over: 0.0309 Episode_Termination/course_exit: 0.5309 Episode_Termination/illegal_nonwheel_contact: 0.0494 ============================================================================== ============================================================================== Learning iteration 2050/50000 Total steps: 8,396,800 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.432s Mean reward: +0.0832 Mean episode length: 1544.7 Actor loss: -8.9509 Critic loss: +0.0014 World model loss: +0.8765 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +55.0149 Episode_Reward/track_angular_velocity: +58.5443 Episode_Reward/upright: +30.1741 Episode_Reward/leg_posture: -5.3506 Episode_Reward/standstill: -8.7029 Episode_Reward/action_rate_l2: -0.1529 Episode_Reward/joint_torques_l2: -0.1602 Episode_Reward/dof_pos_limits: -0.0137 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.2601 Episode_Reward/rolling_leg_motion: -0.9530 Episode_Reward/obstacle_progress: +1.1556 Episode_Reward/wheel_body_intrusion: -0.4590 Episode_Reward/pair_body_intrusion: -0.2495 Curriculum/terrain_level_mean: +0.3486 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.5707 Episode_Termination/fell_over: 0.0157 Episode_Termination/course_exit: 0.3874 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 2100/50000 Total steps: 8,601,600 Steps per second: 7,544 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.433s Mean reward: +0.0833 Mean episode length: 1547.0 Actor loss: -8.9822 Critic loss: +0.0031 World model loss: +0.8908 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +55.0837 Episode_Reward/track_angular_velocity: +58.6544 Episode_Reward/upright: +30.2377 Episode_Reward/leg_posture: -5.2927 Episode_Reward/standstill: -8.6413 Episode_Reward/action_rate_l2: -0.1544 Episode_Reward/joint_torques_l2: -0.1617 Episode_Reward/dof_pos_limits: -0.0121 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.2636 Episode_Reward/rolling_leg_motion: -0.9722 Episode_Reward/obstacle_progress: +1.1478 Episode_Reward/wheel_body_intrusion: -0.4707 Episode_Reward/pair_body_intrusion: -0.2929 Curriculum/terrain_level_mean: +0.3672 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.6195 Episode_Termination/fell_over: 0.0133 Episode_Termination/course_exit: 0.3496 Episode_Termination/illegal_nonwheel_contact: 0.0221 ============================================================================== ============================================================================== Learning iteration 2150/50000 Total steps: 8,806,400 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.110s Learning time: 0.430s Mean reward: +0.0831 Mean episode length: 1618.7 Actor loss: -9.0194 Critic loss: +0.0014 World model loss: +0.9040 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +57.5903 Episode_Reward/track_angular_velocity: +61.3159 Episode_Reward/upright: +31.5922 Episode_Reward/leg_posture: -5.6307 Episode_Reward/standstill: -9.1424 Episode_Reward/action_rate_l2: -0.1658 Episode_Reward/joint_torques_l2: -0.1744 Episode_Reward/dof_pos_limits: -0.0105 Episode_Reward/terminated: -0.0063 Episode_Reward/unnecessary_wheel_lift: -0.2876 Episode_Reward/rolling_leg_motion: -1.0969 Episode_Reward/obstacle_progress: +1.2625 Episode_Reward/wheel_body_intrusion: -0.4594 Episode_Reward/pair_body_intrusion: -0.3407 Curriculum/terrain_level_mean: +0.3865 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.7286 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.2639 Episode_Termination/illegal_nonwheel_contact: 0.0074 ============================================================================== ============================================================================== Learning iteration 2200/50000 Total steps: 9,011,200 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.432s Mean reward: +0.0833 Mean episode length: 1624.9 Actor loss: -9.0414 Critic loss: +0.0020 World model loss: +0.9355 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +57.9081 Episode_Reward/track_angular_velocity: +61.7684 Episode_Reward/upright: +31.7032 Episode_Reward/leg_posture: -5.6941 Episode_Reward/standstill: -9.1897 Episode_Reward/action_rate_l2: -0.1681 Episode_Reward/joint_torques_l2: -0.1784 Episode_Reward/dof_pos_limits: -0.0129 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.3212 Episode_Reward/rolling_leg_motion: -1.1081 Episode_Reward/obstacle_progress: +1.3751 Episode_Reward/wheel_body_intrusion: -0.4374 Episode_Reward/pair_body_intrusion: -0.3114 Curriculum/terrain_level_mean: +0.4080 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1287 Episode_Termination/time_out: 0.7771 Episode_Termination/fell_over: 0.0062 Episode_Termination/course_exit: 0.2074 Episode_Termination/illegal_nonwheel_contact: 0.0093 ============================================================================== ============================================================================== Learning iteration 2250/50000 Total steps: 9,216,000 Steps per second: 7,545 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.432s Mean reward: +0.0836 Mean episode length: 1616.5 Actor loss: -9.0890 Critic loss: +0.0015 World model loss: +0.9217 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +57.5795 Episode_Reward/track_angular_velocity: +61.5090 Episode_Reward/upright: +31.5817 Episode_Reward/leg_posture: -5.4884 Episode_Reward/standstill: -8.8857 Episode_Reward/action_rate_l2: -0.1708 Episode_Reward/joint_torques_l2: -0.1755 Episode_Reward/dof_pos_limits: -0.0087 Episode_Reward/terminated: -0.0253 Episode_Reward/unnecessary_wheel_lift: -0.3363 Episode_Reward/rolling_leg_motion: -1.1126 Episode_Reward/obstacle_progress: +1.4187 Episode_Reward/wheel_body_intrusion: -0.4230 Episode_Reward/pair_body_intrusion: -0.3278 Curriculum/terrain_level_mean: +0.4272 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1271 Episode_Termination/time_out: 0.7968 Episode_Termination/fell_over: 0.0063 Episode_Termination/course_exit: 0.1778 Episode_Termination/illegal_nonwheel_contact: 0.0190 ============================================================================== ============================================================================== Learning iteration 2300/50000 Total steps: 9,420,800 Steps per second: 7,544 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.434s Mean reward: +0.0844 Mean episode length: 1622.2 Actor loss: -9.1185 Critic loss: +0.0016 World model loss: +0.9912 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +58.0972 Episode_Reward/track_angular_velocity: +61.8677 Episode_Reward/upright: +31.7631 Episode_Reward/leg_posture: -5.2421 Episode_Reward/standstill: -8.5908 Episode_Reward/action_rate_l2: -0.1727 Episode_Reward/joint_torques_l2: -0.1738 Episode_Reward/dof_pos_limits: -0.0057 Episode_Reward/terminated: -0.0063 Episode_Reward/unnecessary_wheel_lift: -0.2816 Episode_Reward/rolling_leg_motion: -1.1119 Episode_Reward/obstacle_progress: +1.5173 Episode_Reward/wheel_body_intrusion: -0.4019 Episode_Reward/pair_body_intrusion: -0.3516 Curriculum/terrain_level_mean: +0.4446 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1245 Episode_Termination/time_out: 0.8020 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.1914 Episode_Termination/illegal_nonwheel_contact: 0.0066 ============================================================================== ============================================================================== Learning iteration 2350/50000 Total steps: 9,625,600 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.443s Mean reward: +0.0846 Mean episode length: 1530.1 Actor loss: -9.1465 Critic loss: +0.0015 World model loss: +0.9900 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +54.7679 Episode_Reward/track_angular_velocity: +58.2720 Episode_Reward/upright: +29.9240 Episode_Reward/leg_posture: -4.9095 Episode_Reward/standstill: -7.6767 Episode_Reward/action_rate_l2: -0.1627 Episode_Reward/joint_torques_l2: -0.1660 Episode_Reward/dof_pos_limits: -0.0053 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.2876 Episode_Reward/rolling_leg_motion: -1.0143 Episode_Reward/obstacle_progress: +1.4346 Episode_Reward/wheel_body_intrusion: -0.4103 Episode_Reward/pair_body_intrusion: -0.3203 Curriculum/terrain_level_mean: +0.4614 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1260 Episode_Termination/time_out: 0.6952 Episode_Termination/fell_over: 0.0190 Episode_Termination/course_exit: 0.2714 Episode_Termination/illegal_nonwheel_contact: 0.0190 ============================================================================== ============================================================================== Learning iteration 2400/50000 Total steps: 9,830,400 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.431s Mean reward: +0.0848 Mean episode length: 1540.1 Actor loss: -9.1875 Critic loss: +0.0012 World model loss: +1.0784 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +55.0564 Episode_Reward/track_angular_velocity: +58.6690 Episode_Reward/upright: +30.1129 Episode_Reward/leg_posture: -4.8512 Episode_Reward/standstill: -7.5538 Episode_Reward/action_rate_l2: -0.1667 Episode_Reward/joint_torques_l2: -0.1663 Episode_Reward/dof_pos_limits: -0.0054 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.2847 Episode_Reward/rolling_leg_motion: -1.0279 Episode_Reward/obstacle_progress: +1.5256 Episode_Reward/wheel_body_intrusion: -0.3891 Episode_Reward/pair_body_intrusion: -0.3452 Curriculum/terrain_level_mean: +0.4751 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1258 Episode_Termination/time_out: 0.6751 Episode_Termination/fell_over: 0.0152 Episode_Termination/course_exit: 0.2843 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 2450/50000 Total steps: 10,035,200 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0850 Mean episode length: 1506.4 Actor loss: -9.1901 Critic loss: +0.0018 World model loss: +1.0148 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +53.9993 Episode_Reward/track_angular_velocity: +57.4367 Episode_Reward/upright: +29.4488 Episode_Reward/leg_posture: -4.6378 Episode_Reward/standstill: -7.5435 Episode_Reward/action_rate_l2: -0.1665 Episode_Reward/joint_torques_l2: -0.1627 Episode_Reward/dof_pos_limits: -0.0044 Episode_Reward/terminated: -0.0265 Episode_Reward/unnecessary_wheel_lift: -0.2171 Episode_Reward/rolling_leg_motion: -0.9791 Episode_Reward/obstacle_progress: +1.6044 Episode_Reward/wheel_body_intrusion: -0.3828 Episode_Reward/pair_body_intrusion: -0.3375 Curriculum/terrain_level_mean: +0.4851 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1249 Episode_Termination/time_out: 0.6993 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.2418 Episode_Termination/illegal_nonwheel_contact: 0.0458 ============================================================================== ============================================================================== Learning iteration 2500/50000 Total steps: 10,240,000 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0850 Mean episode length: 1344.9 Actor loss: -9.2209 Critic loss: +0.0015 World model loss: +1.0290 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1264 Episode_Reward/track_angular_velocity: +51.2668 Episode_Reward/upright: +26.3151 Episode_Reward/leg_posture: -4.0820 Episode_Reward/standstill: -6.6966 Episode_Reward/action_rate_l2: -0.1502 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0049 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.2453 Episode_Reward/rolling_leg_motion: -0.8744 Episode_Reward/obstacle_progress: +1.4176 Episode_Reward/wheel_body_intrusion: -0.3414 Episode_Reward/pair_body_intrusion: -0.3099 Curriculum/terrain_level_mean: +0.4978 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1252 Episode_Termination/time_out: 0.5271 Episode_Termination/fell_over: 0.0155 Episode_Termination/course_exit: 0.4109 Episode_Termination/illegal_nonwheel_contact: 0.0465 ============================================================================== ============================================================================== Learning iteration 2550/50000 Total steps: 10,444,800 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.109s Learning time: 0.433s Mean reward: +0.0851 Mean episode length: 1277.7 Actor loss: -9.2467 Critic loss: +0.0020 World model loss: +1.0852 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7291 Episode_Reward/track_angular_velocity: +48.7438 Episode_Reward/upright: +24.9782 Episode_Reward/leg_posture: -3.8931 Episode_Reward/standstill: -6.2253 Episode_Reward/action_rate_l2: -0.1436 Episode_Reward/joint_torques_l2: -0.1368 Episode_Reward/dof_pos_limits: -0.0036 Episode_Reward/terminated: -0.0250 Episode_Reward/unnecessary_wheel_lift: -0.1821 Episode_Reward/rolling_leg_motion: -0.8361 Episode_Reward/obstacle_progress: +1.3359 Episode_Reward/wheel_body_intrusion: -0.3309 Episode_Reward/pair_body_intrusion: -0.2733 Curriculum/terrain_level_mean: +0.5100 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1240 Episode_Termination/time_out: 0.4080 Episode_Termination/fell_over: 0.0560 Episode_Termination/course_exit: 0.5120 Episode_Termination/illegal_nonwheel_contact: 0.0240 ============================================================================== ============================================================================== Learning iteration 2600/50000 Total steps: 10,649,600 Steps per second: 7,542 Iteration time: 0.543s Collection time: 0.109s Learning time: 0.432s Mean reward: +0.0851 Mean episode length: 1284.6 Actor loss: -9.2713 Critic loss: +0.0012 World model loss: +1.0234 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9370 Episode_Reward/track_angular_velocity: +48.9949 Episode_Reward/upright: +25.0965 Episode_Reward/leg_posture: -3.8452 Episode_Reward/standstill: -6.2666 Episode_Reward/action_rate_l2: -0.1453 Episode_Reward/joint_torques_l2: -0.1349 Episode_Reward/dof_pos_limits: -0.0057 Episode_Reward/terminated: -0.0301 Episode_Reward/unnecessary_wheel_lift: -0.2464 Episode_Reward/rolling_leg_motion: -0.8509 Episode_Reward/obstacle_progress: +1.3053 Episode_Reward/wheel_body_intrusion: -0.2916 Episode_Reward/pair_body_intrusion: -0.2519 Curriculum/terrain_level_mean: +0.5181 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1242 Episode_Termination/time_out: 0.4717 Episode_Termination/fell_over: 0.0566 Episode_Termination/course_exit: 0.4151 Episode_Termination/illegal_nonwheel_contact: 0.0755 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[826] ============================================================================== Learning iteration 2650/50000 Total steps: 10,854,400 Steps per second: 7,542 Iteration time: 0.543s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0852 Mean episode length: 1208.8 Actor loss: -9.2650 Critic loss: +0.0030 World model loss: +1.0813 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.3139 Episode_Reward/track_angular_velocity: +46.1130 Episode_Reward/upright: +23.6261 Episode_Reward/leg_posture: -3.6371 Episode_Reward/standstill: -5.8459 Episode_Reward/action_rate_l2: -0.1400 Episode_Reward/joint_torques_l2: -0.1268 Episode_Reward/dof_pos_limits: -0.0038 Episode_Reward/terminated: -0.0118 Episode_Reward/unnecessary_wheel_lift: -0.2079 Episode_Reward/rolling_leg_motion: -0.8001 Episode_Reward/obstacle_progress: +1.2499 Episode_Reward/wheel_body_intrusion: -0.2801 Episode_Reward/pair_body_intrusion: -0.2374 Curriculum/terrain_level_mean: +0.5325 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1240 Episode_Termination/time_out: 0.3302 Episode_Termination/fell_over: 0.0283 Episode_Termination/course_exit: 0.6226 Episode_Termination/illegal_nonwheel_contact: 0.0189 ============================================================================== ============================================================================== Learning iteration 2700/50000 Total steps: 11,059,200 Steps per second: 7,541 Iteration time: 0.543s Collection time: 0.111s Learning time: 0.436s Mean reward: +0.0851 Mean episode length: 1143.5 Actor loss: -9.3081 Critic loss: +0.0024 World model loss: +1.0784 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.0383 Episode_Reward/track_angular_velocity: +43.6499 Episode_Reward/upright: +22.3552 Episode_Reward/leg_posture: -3.4108 Episode_Reward/standstill: -5.7776 Episode_Reward/action_rate_l2: -0.1336 Episode_Reward/joint_torques_l2: -0.1191 Episode_Reward/dof_pos_limits: -0.0035 Episode_Reward/terminated: -0.0290 Episode_Reward/unnecessary_wheel_lift: -0.2033 Episode_Reward/rolling_leg_motion: -0.7706 Episode_Reward/obstacle_progress: +1.1723 Episode_Reward/wheel_body_intrusion: -0.2546 Episode_Reward/pair_body_intrusion: -0.2165 Curriculum/terrain_level_mean: +0.5439 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1233 Episode_Termination/time_out: 0.3056 Episode_Termination/fell_over: 0.0648 Episode_Termination/course_exit: 0.5741 Episode_Termination/illegal_nonwheel_contact: 0.0648 ============================================================================== /lib/python3.11/site-packages/mujoco_warp/_src/io.py:713: UserWarning: MULTICCD is enabled, but the scene contains CCD pairs without multicontact support: [('CYLINDER', 'CYLINDER'), ('CYLINDER', 'BOX')]. At most 1 contact will be generated for these pairs. warnings.warn( +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_002700.mp4 ============================================================================== Learning iteration 2750/50000 Total steps: 11,264,000 Steps per second: 7,504 Iteration time: 0.546s Collection time: 0.108s Learning time: 0.430s Mean reward: +0.0850 Mean episode length: 1072.0 Actor loss: -9.3175 Critic loss: +0.0039 World model loss: +1.1429 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +38.4169 Episode_Reward/track_angular_velocity: +40.9419 Episode_Reward/upright: +20.9601 Episode_Reward/leg_posture: -3.1575 Episode_Reward/standstill: -5.4837 Episode_Reward/action_rate_l2: -0.1273 Episode_Reward/joint_torques_l2: -0.1100 Episode_Reward/dof_pos_limits: -0.0045 Episode_Reward/terminated: -0.0147 Episode_Reward/unnecessary_wheel_lift: -0.1890 Episode_Reward/rolling_leg_motion: -0.7224 Episode_Reward/obstacle_progress: +1.0591 Episode_Reward/wheel_body_intrusion: -0.2381 Episode_Reward/pair_body_intrusion: -0.1932 Curriculum/terrain_level_mean: +0.5562 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1226 Episode_Termination/time_out: 0.1935 Episode_Termination/fell_over: 0.0430 Episode_Termination/course_exit: 0.7312 Episode_Termination/illegal_nonwheel_contact: 0.0430 ============================================================================== ============================================================================== Learning iteration 2800/50000 Total steps: 11,468,800 Steps per second: 7,506 Iteration time: 0.546s Collection time: 0.109s Learning time: 0.430s Mean reward: +0.0851 Mean episode length: 1071.5 Actor loss: -9.3151 Critic loss: +0.0021 World model loss: +1.1425 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +38.4018 Episode_Reward/track_angular_velocity: +40.9503 Episode_Reward/upright: +20.9699 Episode_Reward/leg_posture: -3.1063 Episode_Reward/standstill: -5.3678 Episode_Reward/action_rate_l2: -0.1301 Episode_Reward/joint_torques_l2: -0.1097 Episode_Reward/dof_pos_limits: -0.0030 Episode_Reward/terminated: -0.0293 Episode_Reward/unnecessary_wheel_lift: -0.1710 Episode_Reward/rolling_leg_motion: -0.7331 Episode_Reward/obstacle_progress: +0.9730 Episode_Reward/wheel_body_intrusion: -0.2203 Episode_Reward/pair_body_intrusion: -0.2034 Curriculum/terrain_level_mean: +0.5681 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1219 Episode_Termination/time_out: 0.2143 Episode_Termination/fell_over: 0.0714 Episode_Termination/course_exit: 0.6607 Episode_Termination/illegal_nonwheel_contact: 0.0536 ============================================================================== ============================================================================== Learning iteration 2850/50000 Total steps: 11,673,600 Steps per second: 7,507 Iteration time: 0.546s Collection time: 0.110s Learning time: 0.431s Mean reward: +0.0849 Mean episode length: 1112.2 Actor loss: -9.3435 Critic loss: +0.0012 World model loss: +1.0899 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +39.9053 Episode_Reward/track_angular_velocity: +42.5406 Episode_Reward/upright: +21.7842 Episode_Reward/leg_posture: -3.1585 Episode_Reward/standstill: -5.9630 Episode_Reward/action_rate_l2: -0.1371 Episode_Reward/joint_torques_l2: -0.1116 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0196 Episode_Reward/unnecessary_wheel_lift: -0.1628 Episode_Reward/rolling_leg_motion: -0.7551 Episode_Reward/obstacle_progress: +0.9683 Episode_Reward/wheel_body_intrusion: -0.2294 Episode_Reward/pair_body_intrusion: -0.2228 Curriculum/terrain_level_mean: +0.5808 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1207 Episode_Termination/time_out: 0.2059 Episode_Termination/fell_over: 0.0490 Episode_Termination/course_exit: 0.7059 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 2900/50000 Total steps: 11,878,400 Steps per second: 7,508 Iteration time: 0.546s Collection time: 0.109s Learning time: 0.432s Mean reward: +0.0850 Mean episode length: 1165.5 Actor loss: -9.3323 Critic loss: +0.0024 World model loss: +1.1033 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.7832 Episode_Reward/track_angular_velocity: +44.6140 Episode_Reward/upright: +22.8384 Episode_Reward/leg_posture: -3.2703 Episode_Reward/standstill: -6.1790 Episode_Reward/action_rate_l2: -0.1459 Episode_Reward/joint_torques_l2: -0.1166 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0205 Episode_Reward/unnecessary_wheel_lift: -0.1959 Episode_Reward/rolling_leg_motion: -0.7963 Episode_Reward/obstacle_progress: +1.0242 Episode_Reward/wheel_body_intrusion: -0.2471 Episode_Reward/pair_body_intrusion: -0.2129 Curriculum/terrain_level_mean: +0.5957 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1198 Episode_Termination/time_out: 0.2705 Episode_Termination/fell_over: 0.0492 Episode_Termination/course_exit: 0.6557 Episode_Termination/illegal_nonwheel_contact: 0.0246 ============================================================================== ============================================================================== Learning iteration 2950/50000 Total steps: 12,083,200 Steps per second: 7,509 Iteration time: 0.545s Collection time: 0.109s Learning time: 0.432s Mean reward: +0.0851 Mean episode length: 1159.4 Actor loss: -9.3537 Critic loss: +0.0035 World model loss: +1.1056 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.5591 Episode_Reward/track_angular_velocity: +44.3345 Episode_Reward/upright: +22.7018 Episode_Reward/leg_posture: -3.1951 Episode_Reward/standstill: -6.1720 Episode_Reward/action_rate_l2: -0.1456 Episode_Reward/joint_torques_l2: -0.1160 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0204 Episode_Reward/unnecessary_wheel_lift: -0.1667 Episode_Reward/rolling_leg_motion: -0.7689 Episode_Reward/obstacle_progress: +1.0242 Episode_Reward/wheel_body_intrusion: -0.2460 Episode_Reward/pair_body_intrusion: -0.1797 Curriculum/terrain_level_mean: +0.6101 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1212 Episode_Termination/time_out: 0.3095 Episode_Termination/fell_over: 0.0476 Episode_Termination/course_exit: 0.6190 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 3000/50000 Total steps: 12,288,000 Steps per second: 7,508 Iteration time: 0.546s Collection time: 0.110s Learning time: 0.440s Mean reward: +0.0850 Mean episode length: 1241.1 Actor loss: -9.3621 Critic loss: +0.0026 World model loss: +1.1739 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.5376 Episode_Reward/track_angular_velocity: +47.4011 Episode_Reward/upright: +24.2916 Episode_Reward/leg_posture: -3.4504 Episode_Reward/standstill: -6.6479 Episode_Reward/action_rate_l2: -0.1574 Episode_Reward/joint_torques_l2: -0.1262 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0170 Episode_Reward/unnecessary_wheel_lift: -0.2011 Episode_Reward/rolling_leg_motion: -0.8202 Episode_Reward/obstacle_progress: +1.1740 Episode_Reward/wheel_body_intrusion: -0.2814 Episode_Reward/pair_body_intrusion: -0.1947 Curriculum/terrain_level_mean: +0.6252 Command/command_stage: 0 Command/command_ranges: [[-0.3000, +0.8000], [-0.2000, +0.2000], [-0.5000, +0.5000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1224 Episode_Termination/time_out: 0.3361 Episode_Termination/fell_over: 0.0504 Episode_Termination/course_exit: 0.6050 Episode_Termination/illegal_nonwheel_contact: 0.0084 ============================================================================== ============================================================================== Learning iteration 3050/50000 Total steps: 12,492,800 Steps per second: 7,507 Iteration time: 0.546s Collection time: 0.111s Learning time: 0.437s Mean reward: +0.0851 Mean episode length: 1284.7 Actor loss: -9.3648 Critic loss: +0.0067 World model loss: +1.1011 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9409 Episode_Reward/track_angular_velocity: +49.1132 Episode_Reward/upright: +25.1388 Episode_Reward/leg_posture: -3.5376 Episode_Reward/standstill: -6.7136 Episode_Reward/action_rate_l2: -0.1656 Episode_Reward/joint_torques_l2: -0.1305 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0201 Episode_Reward/unnecessary_wheel_lift: -0.2259 Episode_Reward/rolling_leg_motion: -0.8466 Episode_Reward/obstacle_progress: +1.2098 Episode_Reward/wheel_body_intrusion: -0.2556 Episode_Reward/pair_body_intrusion: -0.1979 Curriculum/terrain_level_mean: +0.6396 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1213 Episode_Termination/time_out: 0.4228 Episode_Termination/fell_over: 0.0407 Episode_Termination/course_exit: 0.5122 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== ============================================================================== Learning iteration 3100/50000 Total steps: 12,697,600 Steps per second: 7,508 Iteration time: 0.546s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0850 Mean episode length: 1238.0 Actor loss: -9.3792 Critic loss: +0.0026 World model loss: +1.1943 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.0247 Episode_Reward/track_angular_velocity: +47.2652 Episode_Reward/upright: +24.1923 Episode_Reward/leg_posture: -3.4207 Episode_Reward/standstill: -6.2759 Episode_Reward/action_rate_l2: -0.1625 Episode_Reward/joint_torques_l2: -0.1264 Episode_Reward/dof_pos_limits: -0.0037 Episode_Reward/terminated: -0.0242 Episode_Reward/unnecessary_wheel_lift: -0.1595 Episode_Reward/rolling_leg_motion: -0.7976 Episode_Reward/obstacle_progress: +1.1973 Episode_Reward/wheel_body_intrusion: -0.2600 Episode_Reward/pair_body_intrusion: -0.1619 Curriculum/terrain_level_mean: +0.6545 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1228 Episode_Termination/time_out: 0.3741 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.5540 Episode_Termination/illegal_nonwheel_contact: 0.0360 ============================================================================== ============================================================================== Learning iteration 3150/50000 Total steps: 12,902,400 Steps per second: 7,509 Iteration time: 0.546s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0845 Mean episode length: 1268.2 Actor loss: -9.3891 Critic loss: +0.0023 World model loss: +1.1008 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8164 Episode_Reward/track_angular_velocity: +48.3917 Episode_Reward/upright: +24.7734 Episode_Reward/leg_posture: -3.5512 Episode_Reward/standstill: -6.6418 Episode_Reward/action_rate_l2: -0.1709 Episode_Reward/joint_torques_l2: -0.1305 Episode_Reward/dof_pos_limits: -0.0028 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.2061 Episode_Reward/rolling_leg_motion: -0.8237 Episode_Reward/obstacle_progress: +1.2207 Episode_Reward/wheel_body_intrusion: -0.2746 Episode_Reward/pair_body_intrusion: -0.2085 Curriculum/terrain_level_mean: +0.6670 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1236 Episode_Termination/time_out: 0.3971 Episode_Termination/fell_over: 0.0515 Episode_Termination/course_exit: 0.5074 Episode_Termination/illegal_nonwheel_contact: 0.0515 ============================================================================== ============================================================================== Learning iteration 3200/50000 Total steps: 13,107,200 Steps per second: 7,509 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0845 Mean episode length: 1254.1 Actor loss: -9.3876 Critic loss: +0.0017 World model loss: +1.1168 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.1409 Episode_Reward/track_angular_velocity: +47.7854 Episode_Reward/upright: +24.4899 Episode_Reward/leg_posture: -3.5113 Episode_Reward/standstill: -6.4398 Episode_Reward/action_rate_l2: -0.1722 Episode_Reward/joint_torques_l2: -0.1290 Episode_Reward/dof_pos_limits: -0.0024 Episode_Reward/terminated: -0.0196 Episode_Reward/unnecessary_wheel_lift: -0.1818 Episode_Reward/rolling_leg_motion: -0.7906 Episode_Reward/obstacle_progress: +1.2483 Episode_Reward/wheel_body_intrusion: -0.2707 Episode_Reward/pair_body_intrusion: -0.1944 Curriculum/terrain_level_mean: +0.6833 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0985 Metrics/base_velocity/error_vel_yaw: 0.1254 Episode_Termination/time_out: 0.3643 Episode_Termination/fell_over: 0.0500 Episode_Termination/course_exit: 0.5786 Episode_Termination/illegal_nonwheel_contact: 0.0143 ============================================================================== ============================================================================== Learning iteration 3250/50000 Total steps: 13,312,000 Steps per second: 7,509 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.434s Mean reward: +0.0843 Mean episode length: 1234.3 Actor loss: -9.3832 Critic loss: +0.0059 World model loss: +1.1497 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.2401 Episode_Reward/track_angular_velocity: +47.0088 Episode_Reward/upright: +24.0689 Episode_Reward/leg_posture: -3.4462 Episode_Reward/standstill: -6.3685 Episode_Reward/action_rate_l2: -0.1722 Episode_Reward/joint_torques_l2: -0.1274 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.2158 Episode_Reward/rolling_leg_motion: -0.7602 Episode_Reward/obstacle_progress: +1.2820 Episode_Reward/wheel_body_intrusion: -0.2447 Episode_Reward/pair_body_intrusion: -0.1643 Curriculum/terrain_level_mean: +0.6973 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1005 Metrics/base_velocity/error_vel_yaw: 0.1260 Episode_Termination/time_out: 0.2712 Episode_Termination/fell_over: 0.0847 Episode_Termination/course_exit: 0.6186 Episode_Termination/illegal_nonwheel_contact: 0.0593 ============================================================================== ============================================================================== Learning iteration 3300/50000 Total steps: 13,516,800 Steps per second: 7,510 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0845 Mean episode length: 1285.1 Actor loss: -9.3781 Critic loss: +0.0065 World model loss: +1.1530 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8314 Episode_Reward/track_angular_velocity: +48.9360 Episode_Reward/upright: +25.0498 Episode_Reward/leg_posture: -3.5901 Episode_Reward/standstill: -6.2992 Episode_Reward/action_rate_l2: -0.1860 Episode_Reward/joint_torques_l2: -0.1339 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0176 Episode_Reward/unnecessary_wheel_lift: -0.1714 Episode_Reward/rolling_leg_motion: -0.7880 Episode_Reward/obstacle_progress: +1.3813 Episode_Reward/wheel_body_intrusion: -0.2772 Episode_Reward/pair_body_intrusion: -0.1853 Curriculum/terrain_level_mean: +0.7097 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1019 Metrics/base_velocity/error_vel_yaw: 0.1259 Episode_Termination/time_out: 0.3500 Episode_Termination/fell_over: 0.0500 Episode_Termination/course_exit: 0.5917 Episode_Termination/illegal_nonwheel_contact: 0.0083 ============================================================================== ============================================================================== Learning iteration 3350/50000 Total steps: 13,721,600 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.430s Mean reward: +0.0838 Mean episode length: 1293.1 Actor loss: -9.4089 Critic loss: +0.0031 World model loss: +1.1708 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8251 Episode_Reward/track_angular_velocity: +49.2230 Episode_Reward/upright: +25.1571 Episode_Reward/leg_posture: -3.6392 Episode_Reward/standstill: -6.7262 Episode_Reward/action_rate_l2: -0.1864 Episode_Reward/joint_torques_l2: -0.1377 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0278 Episode_Reward/unnecessary_wheel_lift: -0.2129 Episode_Reward/rolling_leg_motion: -0.7895 Episode_Reward/obstacle_progress: +1.4256 Episode_Reward/wheel_body_intrusion: -0.3031 Episode_Reward/pair_body_intrusion: -0.2058 Curriculum/terrain_level_mean: +0.7224 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1046 Metrics/base_velocity/error_vel_yaw: 0.1264 Episode_Termination/time_out: 0.3281 Episode_Termination/fell_over: 0.0625 Episode_Termination/course_exit: 0.5859 Episode_Termination/illegal_nonwheel_contact: 0.0234 ============================================================================== ============================================================================== Learning iteration 3400/50000 Total steps: 13,926,400 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0840 Mean episode length: 1243.8 Actor loss: -9.4017 Critic loss: +0.0014 World model loss: +1.1961 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.0281 Episode_Reward/track_angular_velocity: +47.3126 Episode_Reward/upright: +24.1820 Episode_Reward/leg_posture: -3.4295 Episode_Reward/standstill: -6.1931 Episode_Reward/action_rate_l2: -0.1840 Episode_Reward/joint_torques_l2: -0.1309 Episode_Reward/dof_pos_limits: -0.0022 Episode_Reward/terminated: -0.0486 Episode_Reward/unnecessary_wheel_lift: -0.2527 Episode_Reward/rolling_leg_motion: -0.7538 Episode_Reward/obstacle_progress: +1.3719 Episode_Reward/wheel_body_intrusion: -0.2725 Episode_Reward/pair_body_intrusion: -0.2098 Curriculum/terrain_level_mean: +0.7339 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1051 Metrics/base_velocity/error_vel_yaw: 0.1275 Episode_Termination/time_out: 0.2590 Episode_Termination/fell_over: 0.1079 Episode_Termination/course_exit: 0.5971 Episode_Termination/illegal_nonwheel_contact: 0.0432 ============================================================================== ============================================================================== Learning iteration 3450/50000 Total steps: 14,131,200 Steps per second: 7,512 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.435s Mean reward: +0.0845 Mean episode length: 1317.8 Actor loss: -9.3590 Critic loss: +0.0033 World model loss: +1.2701 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6371 Episode_Reward/track_angular_velocity: +50.1975 Episode_Reward/upright: +25.6021 Episode_Reward/leg_posture: -3.6640 Episode_Reward/standstill: -6.0442 Episode_Reward/action_rate_l2: -0.1950 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0021 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.2304 Episode_Reward/rolling_leg_motion: -0.7871 Episode_Reward/obstacle_progress: +1.4512 Episode_Reward/wheel_body_intrusion: -0.2892 Episode_Reward/pair_body_intrusion: -0.1989 Curriculum/terrain_level_mean: +0.7476 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1256 Episode_Termination/time_out: 0.3333 Episode_Termination/fell_over: 0.0684 Episode_Termination/course_exit: 0.5983 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1177] RECOVERY reset 1 invalid environments; first IDs=[2417] ============================================================================== Learning iteration 3500/50000 Total steps: 14,336,000 Steps per second: 7,512 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.433s Mean reward: +0.0844 Mean episode length: 1349.6 Actor loss: -9.3945 Critic loss: +0.0025 World model loss: +1.2176 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6465 Episode_Reward/track_angular_velocity: +51.3757 Episode_Reward/upright: +26.1960 Episode_Reward/leg_posture: -3.7414 Episode_Reward/standstill: -6.1217 Episode_Reward/action_rate_l2: -0.2051 Episode_Reward/joint_torques_l2: -0.1422 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0290 Episode_Reward/unnecessary_wheel_lift: -0.2208 Episode_Reward/rolling_leg_motion: -0.8153 Episode_Reward/obstacle_progress: +1.4332 Episode_Reward/wheel_body_intrusion: -0.2912 Episode_Reward/pair_body_intrusion: -0.2027 Curriculum/terrain_level_mean: +0.7629 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1055 Metrics/base_velocity/error_vel_yaw: 0.1263 Episode_Termination/time_out: 0.4395 Episode_Termination/fell_over: 0.0510 Episode_Termination/course_exit: 0.4904 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 3550/50000 Total steps: 14,540,800 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.440s Mean reward: +0.0843 Mean episode length: 1218.1 Actor loss: -9.3776 Critic loss: +0.0018 World model loss: +1.2504 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.1607 Episode_Reward/track_angular_velocity: +46.3667 Episode_Reward/upright: +23.6487 Episode_Reward/leg_posture: -3.3351 Episode_Reward/standstill: -5.6754 Episode_Reward/action_rate_l2: -0.1930 Episode_Reward/joint_torques_l2: -0.1273 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0357 Episode_Reward/unnecessary_wheel_lift: -0.2338 Episode_Reward/rolling_leg_motion: -0.7520 Episode_Reward/obstacle_progress: +1.3509 Episode_Reward/wheel_body_intrusion: -0.2612 Episode_Reward/pair_body_intrusion: -0.1958 Curriculum/terrain_level_mean: +0.7749 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1050 Metrics/base_velocity/error_vel_yaw: 0.1270 Episode_Termination/time_out: 0.2698 Episode_Termination/fell_over: 0.0952 Episode_Termination/course_exit: 0.6111 Episode_Termination/illegal_nonwheel_contact: 0.0238 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[972] ============================================================================== Learning iteration 3600/50000 Total steps: 14,745,600 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0844 Mean episode length: 1319.2 Actor loss: -9.3585 Critic loss: +0.0023 World model loss: +1.2755 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5573 Episode_Reward/track_angular_velocity: +50.2050 Episode_Reward/upright: +25.6178 Episode_Reward/leg_posture: -3.5727 Episode_Reward/standstill: -6.0291 Episode_Reward/action_rate_l2: -0.2078 Episode_Reward/joint_torques_l2: -0.1379 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0283 Episode_Reward/unnecessary_wheel_lift: -0.2007 Episode_Reward/rolling_leg_motion: -0.7878 Episode_Reward/obstacle_progress: +1.4535 Episode_Reward/wheel_body_intrusion: -0.2904 Episode_Reward/pair_body_intrusion: -0.2051 Curriculum/terrain_level_mean: +0.7854 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1058 Metrics/base_velocity/error_vel_yaw: 0.1269 Episode_Termination/time_out: 0.3871 Episode_Termination/fell_over: 0.0726 Episode_Termination/course_exit: 0.5242 Episode_Termination/illegal_nonwheel_contact: 0.0161 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1701] ============================================================================== Learning iteration 3650/50000 Total steps: 14,950,400 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0841 Mean episode length: 1329.7 Actor loss: -9.3753 Critic loss: +0.0016 World model loss: +1.3304 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7232 Episode_Reward/track_angular_velocity: +50.6264 Episode_Reward/upright: +25.8291 Episode_Reward/leg_posture: -3.6189 Episode_Reward/standstill: -6.2279 Episode_Reward/action_rate_l2: -0.2153 Episode_Reward/joint_torques_l2: -0.1395 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0338 Episode_Reward/unnecessary_wheel_lift: -0.2700 Episode_Reward/rolling_leg_motion: -0.8079 Episode_Reward/obstacle_progress: +1.4059 Episode_Reward/wheel_body_intrusion: -0.2871 Episode_Reward/pair_body_intrusion: -0.2159 Curriculum/terrain_level_mean: +0.7993 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1074 Metrics/base_velocity/error_vel_yaw: 0.1265 Episode_Termination/time_out: 0.3910 Episode_Termination/fell_over: 0.0705 Episode_Termination/course_exit: 0.5256 Episode_Termination/illegal_nonwheel_contact: 0.0128 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1141] ============================================================================== Learning iteration 3700/50000 Total steps: 15,155,200 Steps per second: 7,512 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.433s Mean reward: +0.0843 Mean episode length: 1337.2 Actor loss: -9.3874 Critic loss: +0.0018 World model loss: +1.3059 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1011 Episode_Reward/track_angular_velocity: +50.9307 Episode_Reward/upright: +25.9934 Episode_Reward/leg_posture: -3.5864 Episode_Reward/standstill: -6.2512 Episode_Reward/action_rate_l2: -0.2234 Episode_Reward/joint_torques_l2: -0.1374 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0261 Episode_Reward/unnecessary_wheel_lift: -0.2507 Episode_Reward/rolling_leg_motion: -0.8145 Episode_Reward/obstacle_progress: +1.4396 Episode_Reward/wheel_body_intrusion: -0.2796 Episode_Reward/pair_body_intrusion: -0.2006 Curriculum/terrain_level_mean: +0.8162 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1060 Metrics/base_velocity/error_vel_yaw: 0.1261 Episode_Termination/time_out: 0.4024 Episode_Termination/fell_over: 0.0592 Episode_Termination/course_exit: 0.5385 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2850] RECOVERY reset 1 invalid environments; first IDs=[2386] ============================================================================== Learning iteration 3750/50000 Total steps: 15,360,000 Steps per second: 7,512 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.435s Mean reward: +0.0844 Mean episode length: 1416.2 Actor loss: -9.3419 Critic loss: +0.0036 World model loss: +1.3088 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6998 Episode_Reward/track_angular_velocity: +53.9264 Episode_Reward/upright: +27.5333 Episode_Reward/leg_posture: -3.7673 Episode_Reward/standstill: -6.3067 Episode_Reward/action_rate_l2: -0.2405 Episode_Reward/joint_torques_l2: -0.1461 Episode_Reward/dof_pos_limits: -0.0020 Episode_Reward/terminated: -0.0249 Episode_Reward/unnecessary_wheel_lift: -0.2548 Episode_Reward/rolling_leg_motion: -0.8567 Episode_Reward/obstacle_progress: +1.4905 Episode_Reward/wheel_body_intrusion: -0.2962 Episode_Reward/pair_body_intrusion: -0.2196 Curriculum/terrain_level_mean: +0.8289 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1074 Metrics/base_velocity/error_vel_yaw: 0.1260 Episode_Termination/time_out: 0.5490 Episode_Termination/fell_over: 0.0392 Episode_Termination/course_exit: 0.3922 Episode_Termination/illegal_nonwheel_contact: 0.0196 ============================================================================== ============================================================================== Learning iteration 3800/50000 Total steps: 15,564,800 Steps per second: 7,512 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0847 Mean episode length: 1413.8 Actor loss: -9.3776 Critic loss: +0.0042 World model loss: +1.2664 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6175 Episode_Reward/track_angular_velocity: +53.8636 Episode_Reward/upright: +27.5116 Episode_Reward/leg_posture: -3.6954 Episode_Reward/standstill: -6.1607 Episode_Reward/action_rate_l2: -0.2477 Episode_Reward/joint_torques_l2: -0.1446 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.2056 Episode_Reward/rolling_leg_motion: -0.8620 Episode_Reward/obstacle_progress: +1.5227 Episode_Reward/wheel_body_intrusion: -0.2761 Episode_Reward/pair_body_intrusion: -0.2289 Curriculum/terrain_level_mean: +0.8442 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1072 Metrics/base_velocity/error_vel_yaw: 0.1262 Episode_Termination/time_out: 0.5196 Episode_Termination/fell_over: 0.0223 Episode_Termination/course_exit: 0.4469 Episode_Termination/illegal_nonwheel_contact: 0.0112 ============================================================================== ============================================================================== Learning iteration 3850/50000 Total steps: 15,769,600 Steps per second: 7,513 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.429s Mean reward: +0.0847 Mean episode length: 1478.3 Actor loss: -9.3474 Critic loss: +0.0040 World model loss: +1.2753 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7782 Episode_Reward/track_angular_velocity: +56.2916 Episode_Reward/upright: +28.7531 Episode_Reward/leg_posture: -3.8313 Episode_Reward/standstill: -6.2708 Episode_Reward/action_rate_l2: -0.2588 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0202 Episode_Reward/unnecessary_wheel_lift: -0.2240 Episode_Reward/rolling_leg_motion: -0.8986 Episode_Reward/obstacle_progress: +1.6039 Episode_Reward/wheel_body_intrusion: -0.2965 Episode_Reward/pair_body_intrusion: -0.2250 Curriculum/terrain_level_mean: +0.8574 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1075 Metrics/base_velocity/error_vel_yaw: 0.1265 Episode_Termination/time_out: 0.6126 Episode_Termination/fell_over: 0.0366 Episode_Termination/course_exit: 0.3508 Episode_Termination/illegal_nonwheel_contact: 0.0052 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2261] ============================================================================== Learning iteration 3900/50000 Total steps: 15,974,400 Steps per second: 7,513 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.434s Mean reward: +0.0847 Mean episode length: 1490.2 Actor loss: -9.3798 Critic loss: +0.0015 World model loss: +1.3116 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.2759 Episode_Reward/track_angular_velocity: +56.7564 Episode_Reward/upright: +28.9578 Episode_Reward/leg_posture: -3.9387 Episode_Reward/standstill: -6.2940 Episode_Reward/action_rate_l2: -0.2643 Episode_Reward/joint_torques_l2: -0.1549 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0349 Episode_Reward/unnecessary_wheel_lift: -0.2264 Episode_Reward/rolling_leg_motion: -0.9287 Episode_Reward/obstacle_progress: +1.6721 Episode_Reward/wheel_body_intrusion: -0.2893 Episode_Reward/pair_body_intrusion: -0.2548 Curriculum/terrain_level_mean: +0.8708 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1071 Metrics/base_velocity/error_vel_yaw: 0.1261 Episode_Termination/time_out: 0.6360 Episode_Termination/fell_over: 0.0526 Episode_Termination/course_exit: 0.3114 Episode_Termination/illegal_nonwheel_contact: 0.0088 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[529] ============================================================================== Learning iteration 3950/50000 Total steps: 16,179,200 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.432s Mean reward: +0.0846 Mean episode length: 1545.9 Actor loss: -9.3422 Critic loss: +0.0027 World model loss: +1.3565 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +53.1013 Episode_Reward/track_angular_velocity: +58.8831 Episode_Reward/upright: +30.0217 Episode_Reward/leg_posture: -4.1239 Episode_Reward/standstill: -6.6178 Episode_Reward/action_rate_l2: -0.2777 Episode_Reward/joint_torques_l2: -0.1626 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.2465 Episode_Reward/rolling_leg_motion: -0.9708 Episode_Reward/obstacle_progress: +1.7566 Episode_Reward/wheel_body_intrusion: -0.2754 Episode_Reward/pair_body_intrusion: -0.2543 Curriculum/terrain_level_mean: +0.8840 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1077 Metrics/base_velocity/error_vel_yaw: 0.1260 Episode_Termination/time_out: 0.7395 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.2375 Episode_Termination/illegal_nonwheel_contact: 0.0038 ============================================================================== ============================================================================== Learning iteration 4000/50000 Total steps: 16,384,000 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0845 Mean episode length: 1493.5 Actor loss: -9.3541 Critic loss: +0.0038 World model loss: +1.3027 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.2340 Episode_Reward/track_angular_velocity: +56.8683 Episode_Reward/upright: +29.0014 Episode_Reward/leg_posture: -4.0629 Episode_Reward/standstill: -6.3987 Episode_Reward/action_rate_l2: -0.2737 Episode_Reward/joint_torques_l2: -0.1591 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0204 Episode_Reward/unnecessary_wheel_lift: -0.2504 Episode_Reward/rolling_leg_motion: -0.9498 Episode_Reward/obstacle_progress: +1.6672 Episode_Reward/wheel_body_intrusion: -0.3001 Episode_Reward/pair_body_intrusion: -0.2304 Curriculum/terrain_level_mean: +0.9001 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1080 Metrics/base_velocity/error_vel_yaw: 0.1263 Episode_Termination/time_out: 0.6806 Episode_Termination/fell_over: 0.0266 Episode_Termination/course_exit: 0.2928 Episode_Termination/illegal_nonwheel_contact: 0.0076 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[820] ============================================================================== Learning iteration 4050/50000 Total steps: 16,588,800 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.435s Mean reward: +0.0844 Mean episode length: 1481.8 Actor loss: -9.3879 Critic loss: +0.0038 World model loss: +1.3393 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7850 Episode_Reward/track_angular_velocity: +56.3318 Episode_Reward/upright: +28.7533 Episode_Reward/leg_posture: -3.9532 Episode_Reward/standstill: -6.3909 Episode_Reward/action_rate_l2: -0.2765 Episode_Reward/joint_torques_l2: -0.1568 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0289 Episode_Reward/unnecessary_wheel_lift: -0.2617 Episode_Reward/rolling_leg_motion: -0.9321 Episode_Reward/obstacle_progress: +1.6833 Episode_Reward/wheel_body_intrusion: -0.2893 Episode_Reward/pair_body_intrusion: -0.2070 Curriculum/terrain_level_mean: +0.9106 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1083 Metrics/base_velocity/error_vel_yaw: 0.1284 Episode_Termination/time_out: 0.6883 Episode_Termination/fell_over: 0.0433 Episode_Termination/course_exit: 0.2684 Episode_Termination/illegal_nonwheel_contact: 0.0043 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2101] ============================================================================== Learning iteration 4100/50000 Total steps: 16,793,600 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.435s Mean reward: +0.0844 Mean episode length: 1380.8 Actor loss: -9.3672 Critic loss: +0.0070 World model loss: +1.3293 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2577 Episode_Reward/track_angular_velocity: +52.5476 Episode_Reward/upright: +26.8008 Episode_Reward/leg_posture: -3.7190 Episode_Reward/standstill: -5.9808 Episode_Reward/action_rate_l2: -0.2649 Episode_Reward/joint_torques_l2: -0.1477 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1929 Episode_Reward/rolling_leg_motion: -0.8529 Episode_Reward/obstacle_progress: +1.6489 Episode_Reward/wheel_body_intrusion: -0.2940 Episode_Reward/pair_body_intrusion: -0.2317 Curriculum/terrain_level_mean: +0.9248 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1092 Metrics/base_velocity/error_vel_yaw: 0.1274 Episode_Termination/time_out: 0.5600 Episode_Termination/fell_over: 0.0343 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0057 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2190] ============================================================================== Learning iteration 4150/50000 Total steps: 16,998,400 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.434s Mean reward: +0.0840 Mean episode length: 1387.9 Actor loss: -9.3751 Critic loss: +0.0084 World model loss: +1.3369 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4837 Episode_Reward/track_angular_velocity: +52.7422 Episode_Reward/upright: +26.9335 Episode_Reward/leg_posture: -3.7494 Episode_Reward/standstill: -6.3825 Episode_Reward/action_rate_l2: -0.2770 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0025 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.2123 Episode_Reward/rolling_leg_motion: -0.8678 Episode_Reward/obstacle_progress: +1.6323 Episode_Reward/wheel_body_intrusion: -0.2839 Episode_Reward/pair_body_intrusion: -0.2476 Curriculum/terrain_level_mean: +0.9370 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1092 Metrics/base_velocity/error_vel_yaw: 0.1292 Episode_Termination/time_out: 0.5858 Episode_Termination/fell_over: 0.0473 Episode_Termination/course_exit: 0.3669 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4200/50000 Total steps: 17,203,200 Steps per second: 7,514 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.435s Mean reward: +0.0840 Mean episode length: 1375.1 Actor loss: -9.3389 Critic loss: +0.0115 World model loss: +1.3954 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1619 Episode_Reward/track_angular_velocity: +52.2014 Episode_Reward/upright: +26.6942 Episode_Reward/leg_posture: -3.6622 Episode_Reward/standstill: -6.4330 Episode_Reward/action_rate_l2: -0.3006 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0403 Episode_Reward/unnecessary_wheel_lift: -0.2367 Episode_Reward/rolling_leg_motion: -0.8847 Episode_Reward/obstacle_progress: +1.6961 Episode_Reward/wheel_body_intrusion: -0.2699 Episode_Reward/pair_body_intrusion: -0.2179 Curriculum/terrain_level_mean: +0.9456 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1083 Metrics/base_velocity/error_vel_yaw: 0.1310 Episode_Termination/time_out: 0.5541 Episode_Termination/fell_over: 0.0828 Episode_Termination/course_exit: 0.3503 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 4250/50000 Total steps: 17,408,000 Steps per second: 7,513 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.439s Mean reward: +0.0838 Mean episode length: 1314.4 Actor loss: -9.3490 Critic loss: +0.0109 World model loss: +1.3547 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.0273 Episode_Reward/track_angular_velocity: +49.8287 Episode_Reward/upright: +25.4951 Episode_Reward/leg_posture: -3.5172 Episode_Reward/standstill: -6.1885 Episode_Reward/action_rate_l2: -0.3030 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.2188 Episode_Reward/rolling_leg_motion: -0.8675 Episode_Reward/obstacle_progress: +1.5794 Episode_Reward/wheel_body_intrusion: -0.2755 Episode_Reward/pair_body_intrusion: -0.2156 Curriculum/terrain_level_mean: +0.9597 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1091 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4258 Episode_Termination/fell_over: 0.0581 Episode_Termination/course_exit: 0.5097 Episode_Termination/illegal_nonwheel_contact: 0.0065 ============================================================================== ============================================================================== Learning iteration 4300/50000 Total steps: 17,612,800 Steps per second: 7,513 Iteration time: 0.545s Collection time: 0.111s Learning time: 0.436s Mean reward: +0.0840 Mean episode length: 1216.9 Actor loss: -9.3960 Critic loss: +0.0025 World model loss: +1.4374 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.7825 Episode_Reward/track_angular_velocity: +46.1770 Episode_Reward/upright: +23.5933 Episode_Reward/leg_posture: -3.2030 Episode_Reward/standstill: -5.7822 Episode_Reward/action_rate_l2: -0.2856 Episode_Reward/joint_torques_l2: -0.1317 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0309 Episode_Reward/unnecessary_wheel_lift: -0.1381 Episode_Reward/rolling_leg_motion: -0.7995 Episode_Reward/obstacle_progress: +1.4615 Episode_Reward/wheel_body_intrusion: -0.2296 Episode_Reward/pair_body_intrusion: -0.1642 Curriculum/terrain_level_mean: +0.9688 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1082 Metrics/base_velocity/error_vel_yaw: 0.1312 Episode_Termination/time_out: 0.4297 Episode_Termination/fell_over: 0.0938 Episode_Termination/course_exit: 0.4688 Episode_Termination/illegal_nonwheel_contact: 0.0078 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[802] ============================================================================== Learning iteration 4350/50000 Total steps: 17,817,600 Steps per second: 7,511 Iteration time: 0.545s Collection time: 0.112s Learning time: 0.447s Mean reward: +0.0844 Mean episode length: 1258.7 Actor loss: -9.4101 Critic loss: +0.0026 World model loss: +1.3852 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.3230 Episode_Reward/track_angular_velocity: +47.7322 Episode_Reward/upright: +24.4470 Episode_Reward/leg_posture: -3.1841 Episode_Reward/standstill: -5.8533 Episode_Reward/action_rate_l2: -0.3053 Episode_Reward/joint_torques_l2: -0.1341 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1445 Episode_Reward/rolling_leg_motion: -0.8117 Episode_Reward/obstacle_progress: +1.5533 Episode_Reward/wheel_body_intrusion: -0.2403 Episode_Reward/pair_body_intrusion: -0.1622 Curriculum/terrain_level_mean: +0.9817 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1070 Metrics/base_velocity/error_vel_yaw: 0.1321 Episode_Termination/time_out: 0.4351 Episode_Termination/fell_over: 0.0534 Episode_Termination/course_exit: 0.4962 Episode_Termination/illegal_nonwheel_contact: 0.0229 ============================================================================== ============================================================================== Learning iteration 4400/50000 Total steps: 18,022,400 Steps per second: 7,510 Iteration time: 0.545s Collection time: 0.110s Learning time: 0.437s Mean reward: +0.0842 Mean episode length: 1202.5 Actor loss: -9.4081 Critic loss: +0.0034 World model loss: +1.4562 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.3537 Episode_Reward/track_angular_velocity: +45.5888 Episode_Reward/upright: +23.3546 Episode_Reward/leg_posture: -3.0723 Episode_Reward/standstill: -5.6774 Episode_Reward/action_rate_l2: -0.3032 Episode_Reward/joint_torques_l2: -0.1283 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0258 Episode_Reward/unnecessary_wheel_lift: -0.1311 Episode_Reward/rolling_leg_motion: -0.7913 Episode_Reward/obstacle_progress: +1.4555 Episode_Reward/wheel_body_intrusion: -0.2423 Episode_Reward/pair_body_intrusion: -0.1681 Curriculum/terrain_level_mean: +0.9939 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1074 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.3577 Episode_Termination/fell_over: 0.0511 Episode_Termination/course_exit: 0.5620 Episode_Termination/illegal_nonwheel_contact: 0.0292 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1560] ============================================================================== Learning iteration 4450/50000 Total steps: 18,227,200 Steps per second: 7,516 Iteration time: 0.545s Collection time: 0.103s Learning time: 0.407s Mean reward: +0.0844 Mean episode length: 1244.1 Actor loss: -9.3731 Critic loss: +0.0049 World model loss: +1.4871 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.8766 Episode_Reward/track_angular_velocity: +47.2121 Episode_Reward/upright: +24.1750 Episode_Reward/leg_posture: -3.1762 Episode_Reward/standstill: -5.7684 Episode_Reward/action_rate_l2: -0.3155 Episode_Reward/joint_torques_l2: -0.1331 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0146 Episode_Reward/unnecessary_wheel_lift: -0.1353 Episode_Reward/rolling_leg_motion: -0.7980 Episode_Reward/obstacle_progress: +1.5198 Episode_Reward/wheel_body_intrusion: -0.2412 Episode_Reward/pair_body_intrusion: -0.1697 Curriculum/terrain_level_mean: +1.0054 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1065 Metrics/base_velocity/error_vel_yaw: 0.1309 Episode_Termination/time_out: 0.4375 Episode_Termination/fell_over: 0.0391 Episode_Termination/course_exit: 0.5156 Episode_Termination/illegal_nonwheel_contact: 0.0234 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[424] ============================================================================== Learning iteration 4500/50000 Total steps: 18,432,000 Steps per second: 7,522 Iteration time: 0.545s Collection time: 0.102s Learning time: 0.403s Mean reward: +0.0840 Mean episode length: 1194.4 Actor loss: -9.3845 Critic loss: +0.0036 World model loss: +1.4965 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.1738 Episode_Reward/track_angular_velocity: +45.3390 Episode_Reward/upright: +23.1956 Episode_Reward/leg_posture: -3.0665 Episode_Reward/standstill: -5.8966 Episode_Reward/action_rate_l2: -0.3113 Episode_Reward/joint_torques_l2: -0.1284 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0327 Episode_Reward/unnecessary_wheel_lift: -0.1520 Episode_Reward/rolling_leg_motion: -0.7648 Episode_Reward/obstacle_progress: +1.4283 Episode_Reward/wheel_body_intrusion: -0.2313 Episode_Reward/pair_body_intrusion: -0.1820 Curriculum/terrain_level_mean: +1.0166 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1063 Metrics/base_velocity/error_vel_yaw: 0.1309 Episode_Termination/time_out: 0.3223 Episode_Termination/fell_over: 0.1157 Episode_Termination/course_exit: 0.5620 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4550/50000 Total steps: 18,636,800 Steps per second: 7,533 Iteration time: 0.544s Collection time: 0.098s Learning time: 0.373s Mean reward: +0.0836 Mean episode length: 1280.8 Actor loss: -9.4140 Critic loss: +0.0032 World model loss: +1.4973 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.0624 Episode_Reward/track_angular_velocity: +48.5585 Episode_Reward/upright: +24.8761 Episode_Reward/leg_posture: -3.3196 Episode_Reward/standstill: -6.7015 Episode_Reward/action_rate_l2: -0.3353 Episode_Reward/joint_torques_l2: -0.1376 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0350 Episode_Reward/unnecessary_wheel_lift: -0.1875 Episode_Reward/rolling_leg_motion: -0.8225 Episode_Reward/obstacle_progress: +1.5289 Episode_Reward/wheel_body_intrusion: -0.2521 Episode_Reward/pair_body_intrusion: -0.1937 Curriculum/terrain_level_mean: +1.0273 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1075 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.3926 Episode_Termination/fell_over: 0.1037 Episode_Termination/course_exit: 0.5037 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4600/50000 Total steps: 18,841,600 Steps per second: 7,574 Iteration time: 0.541s Collection time: 0.077s Learning time: 0.194s Mean reward: +0.0840 Mean episode length: 1213.6 Actor loss: -9.4142 Critic loss: +0.0025 World model loss: +1.4365 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.7799 Episode_Reward/track_angular_velocity: +46.0153 Episode_Reward/upright: +23.6022 Episode_Reward/leg_posture: -3.0970 Episode_Reward/standstill: -5.9122 Episode_Reward/action_rate_l2: -0.3165 Episode_Reward/joint_torques_l2: -0.1313 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.2000 Episode_Reward/rolling_leg_motion: -0.7685 Episode_Reward/obstacle_progress: +1.4348 Episode_Reward/wheel_body_intrusion: -0.2546 Episode_Reward/pair_body_intrusion: -0.1788 Curriculum/terrain_level_mean: +1.0388 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1067 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.3778 Episode_Termination/fell_over: 0.0741 Episode_Termination/course_exit: 0.5333 Episode_Termination/illegal_nonwheel_contact: 0.0148 ============================================================================== ============================================================================== Learning iteration 4650/50000 Total steps: 19,046,400 Steps per second: 7,579 Iteration time: 0.540s Collection time: 0.102s Learning time: 0.405s Mean reward: +0.0839 Mean episode length: 1267.4 Actor loss: -9.4090 Critic loss: +0.0033 World model loss: +1.4178 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.6214 Episode_Reward/track_angular_velocity: +48.0383 Episode_Reward/upright: +24.6476 Episode_Reward/leg_posture: -3.2162 Episode_Reward/standstill: -6.3863 Episode_Reward/action_rate_l2: -0.3468 Episode_Reward/joint_torques_l2: -0.1372 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.2043 Episode_Reward/rolling_leg_motion: -0.8162 Episode_Reward/obstacle_progress: +1.5446 Episode_Reward/wheel_body_intrusion: -0.2550 Episode_Reward/pair_body_intrusion: -0.1644 Curriculum/terrain_level_mean: +1.0471 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1071 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4597 Episode_Termination/fell_over: 0.0968 Episode_Termination/course_exit: 0.4355 Episode_Termination/illegal_nonwheel_contact: 0.0081 ============================================================================== ============================================================================== Learning iteration 4700/50000 Total steps: 19,251,200 Steps per second: 7,584 Iteration time: 0.540s Collection time: 0.102s Learning time: 0.406s Mean reward: +0.0841 Mean episode length: 1305.1 Actor loss: -9.4048 Critic loss: +0.0026 World model loss: +1.4708 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8979 Episode_Reward/track_angular_velocity: +49.4907 Episode_Reward/upright: +25.4037 Episode_Reward/leg_posture: -3.2582 Episode_Reward/standstill: -6.3748 Episode_Reward/action_rate_l2: -0.3527 Episode_Reward/joint_torques_l2: -0.1412 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0178 Episode_Reward/unnecessary_wheel_lift: -0.1196 Episode_Reward/rolling_leg_motion: -0.8267 Episode_Reward/obstacle_progress: +1.5270 Episode_Reward/wheel_body_intrusion: -0.2552 Episode_Reward/pair_body_intrusion: -0.2066 Curriculum/terrain_level_mean: +1.0586 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1070 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4254 Episode_Termination/fell_over: 0.0522 Episode_Termination/course_exit: 0.5299 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4750/50000 Total steps: 19,456,000 Steps per second: 7,589 Iteration time: 0.540s Collection time: 0.102s Learning time: 0.406s Mean reward: +0.0840 Mean episode length: 1220.6 Actor loss: -9.4095 Critic loss: +0.0060 World model loss: +1.4412 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.0473 Episode_Reward/track_angular_velocity: +46.2478 Episode_Reward/upright: +23.7440 Episode_Reward/leg_posture: -3.1106 Episode_Reward/standstill: -5.9072 Episode_Reward/action_rate_l2: -0.3415 Episode_Reward/joint_torques_l2: -0.1337 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0429 Episode_Reward/unnecessary_wheel_lift: -0.1891 Episode_Reward/rolling_leg_motion: -0.7728 Episode_Reward/obstacle_progress: +1.4767 Episode_Reward/wheel_body_intrusion: -0.2490 Episode_Reward/pair_body_intrusion: -0.2111 Curriculum/terrain_level_mean: +1.0667 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1063 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.3636 Episode_Termination/fell_over: 0.1405 Episode_Termination/course_exit: 0.4876 Episode_Termination/illegal_nonwheel_contact: 0.0083 ============================================================================== ============================================================================== Learning iteration 4800/50000 Total steps: 19,660,800 Steps per second: 7,593 Iteration time: 0.539s Collection time: 0.103s Learning time: 0.406s Mean reward: +0.0841 Mean episode length: 1246.1 Actor loss: -9.4108 Critic loss: +0.0054 World model loss: +1.3881 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.8039 Episode_Reward/track_angular_velocity: +47.2139 Episode_Reward/upright: +24.2650 Episode_Reward/leg_posture: -3.1401 Episode_Reward/standstill: -5.9095 Episode_Reward/action_rate_l2: -0.3511 Episode_Reward/joint_torques_l2: -0.1353 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0365 Episode_Reward/unnecessary_wheel_lift: -0.2086 Episode_Reward/rolling_leg_motion: -0.7882 Episode_Reward/obstacle_progress: +1.5237 Episode_Reward/wheel_body_intrusion: -0.2510 Episode_Reward/pair_body_intrusion: -0.1805 Curriculum/terrain_level_mean: +1.0796 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1072 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.3660 Episode_Termination/fell_over: 0.0980 Episode_Termination/course_exit: 0.5359 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4850/50000 Total steps: 19,865,600 Steps per second: 7,598 Iteration time: 0.539s Collection time: 0.102s Learning time: 0.406s Mean reward: +0.0846 Mean episode length: 1272.6 Actor loss: -9.4127 Critic loss: +0.0023 World model loss: +1.4552 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.8437 Episode_Reward/track_angular_velocity: +48.2440 Episode_Reward/upright: +24.8163 Episode_Reward/leg_posture: -3.1715 Episode_Reward/standstill: -5.7300 Episode_Reward/action_rate_l2: -0.3622 Episode_Reward/joint_torques_l2: -0.1394 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0224 Episode_Reward/unnecessary_wheel_lift: -0.1419 Episode_Reward/rolling_leg_motion: -0.7989 Episode_Reward/obstacle_progress: +1.5208 Episode_Reward/wheel_body_intrusion: -0.2600 Episode_Reward/pair_body_intrusion: -0.1872 Curriculum/terrain_level_mean: +1.0913 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1062 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4265 Episode_Termination/fell_over: 0.0662 Episode_Termination/course_exit: 0.5074 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 4900/50000 Total steps: 20,070,400 Steps per second: 7,600 Iteration time: 0.539s Collection time: 0.106s Learning time: 0.415s Mean reward: +0.0840 Mean episode length: 1265.0 Actor loss: -9.3824 Critic loss: +0.0050 World model loss: +1.3694 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.5407 Episode_Reward/track_angular_velocity: +47.9287 Episode_Reward/upright: +24.6061 Episode_Reward/leg_posture: -3.1402 Episode_Reward/standstill: -6.2414 Episode_Reward/action_rate_l2: -0.3535 Episode_Reward/joint_torques_l2: -0.1390 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0420 Episode_Reward/unnecessary_wheel_lift: -0.2304 Episode_Reward/rolling_leg_motion: -0.7937 Episode_Reward/obstacle_progress: +1.5242 Episode_Reward/wheel_body_intrusion: -0.2430 Episode_Reward/pair_body_intrusion: -0.1817 Curriculum/terrain_level_mean: +1.1030 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1067 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4255 Episode_Termination/fell_over: 0.1135 Episode_Termination/course_exit: 0.4539 Episode_Termination/illegal_nonwheel_contact: 0.0071 ============================================================================== ============================================================================== Learning iteration 4950/50000 Total steps: 20,275,200 Steps per second: 7,601 Iteration time: 0.539s Collection time: 0.109s Learning time: 0.427s Mean reward: +0.0847 Mean episode length: 1300.8 Actor loss: -9.4156 Critic loss: +0.0028 World model loss: +1.4785 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8555 Episode_Reward/track_angular_velocity: +49.3052 Episode_Reward/upright: +25.3262 Episode_Reward/leg_posture: -3.1809 Episode_Reward/standstill: -5.8020 Episode_Reward/action_rate_l2: -0.3718 Episode_Reward/joint_torques_l2: -0.1432 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0178 Episode_Reward/unnecessary_wheel_lift: -0.1549 Episode_Reward/rolling_leg_motion: -0.8103 Episode_Reward/obstacle_progress: +1.6366 Episode_Reward/wheel_body_intrusion: -0.2577 Episode_Reward/pair_body_intrusion: -0.1705 Curriculum/terrain_level_mean: +1.1130 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1059 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4733 Episode_Termination/fell_over: 0.0534 Episode_Termination/course_exit: 0.4733 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 5000/50000 Total steps: 20,480,000 Steps per second: 7,602 Iteration time: 0.539s Collection time: 0.107s Learning time: 0.422s Mean reward: +0.0846 Mean episode length: 1252.6 Actor loss: -9.4025 Critic loss: +0.0030 World model loss: +1.4712 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.2498 Episode_Reward/track_angular_velocity: +47.4359 Episode_Reward/upright: +24.3571 Episode_Reward/leg_posture: -3.0467 Episode_Reward/standstill: -5.7600 Episode_Reward/action_rate_l2: -0.3618 Episode_Reward/joint_torques_l2: -0.1424 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0294 Episode_Reward/unnecessary_wheel_lift: -0.1330 Episode_Reward/rolling_leg_motion: -0.7740 Episode_Reward/obstacle_progress: +1.6314 Episode_Reward/wheel_body_intrusion: -0.2408 Episode_Reward/pair_body_intrusion: -0.1839 Curriculum/terrain_level_mean: +1.1240 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1056 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.3485 Episode_Termination/fell_over: 0.0909 Episode_Termination/course_exit: 0.5606 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2643] RECOVERY reset 1 invalid environments; first IDs=[1752] ============================================================================== Learning iteration 5050/50000 Total steps: 20,684,800 Steps per second: 7,605 Iteration time: 0.539s Collection time: 0.108s Learning time: 0.409s Mean reward: +0.0845 Mean episode length: 1249.0 Actor loss: -9.3800 Critic loss: +0.0038 World model loss: +1.4425 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.1158 Episode_Reward/track_angular_velocity: +47.3319 Episode_Reward/upright: +24.2410 Episode_Reward/leg_posture: -3.0729 Episode_Reward/standstill: -5.9413 Episode_Reward/action_rate_l2: -0.3630 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0073 Episode_Reward/unnecessary_wheel_lift: -0.1218 Episode_Reward/rolling_leg_motion: -0.7558 Episode_Reward/obstacle_progress: +1.6815 Episode_Reward/wheel_body_intrusion: -0.2501 Episode_Reward/pair_body_intrusion: -0.1852 Curriculum/terrain_level_mean: +1.1377 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1058 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.3667 Episode_Termination/fell_over: 0.0250 Episode_Termination/course_exit: 0.6083 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[855] ============================================================================== Learning iteration 5100/50000 Total steps: 20,889,600 Steps per second: 7,608 Iteration time: 0.538s Collection time: 0.104s Learning time: 0.410s Mean reward: +0.0844 Mean episode length: 1317.5 Actor loss: -9.4311 Critic loss: +0.0029 World model loss: +1.4028 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5144 Episode_Reward/track_angular_velocity: +49.8937 Episode_Reward/upright: +25.5575 Episode_Reward/leg_posture: -3.2321 Episode_Reward/standstill: -6.3496 Episode_Reward/action_rate_l2: -0.3851 Episode_Reward/joint_torques_l2: -0.1538 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0309 Episode_Reward/unnecessary_wheel_lift: -0.1324 Episode_Reward/rolling_leg_motion: -0.7983 Episode_Reward/obstacle_progress: +1.7543 Episode_Reward/wheel_body_intrusion: -0.2676 Episode_Reward/pair_body_intrusion: -0.2218 Curriculum/terrain_level_mean: +1.1477 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1054 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4091 Episode_Termination/fell_over: 0.0833 Episode_Termination/course_exit: 0.5076 Episode_Termination/illegal_nonwheel_contact: 0.0076 ============================================================================== ============================================================================== Learning iteration 5150/50000 Total steps: 21,094,400 Steps per second: 7,631 Iteration time: 0.537s Collection time: 0.089s Learning time: 0.287s Mean reward: +0.0845 Mean episode length: 1246.7 Actor loss: -9.4228 Critic loss: +0.0028 World model loss: +1.3794 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.1903 Episode_Reward/track_angular_velocity: +47.2656 Episode_Reward/upright: +24.1829 Episode_Reward/leg_posture: -2.9714 Episode_Reward/standstill: -6.0972 Episode_Reward/action_rate_l2: -0.3700 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0365 Episode_Reward/unnecessary_wheel_lift: -0.1551 Episode_Reward/rolling_leg_motion: -0.7545 Episode_Reward/obstacle_progress: +1.6768 Episode_Reward/wheel_body_intrusion: -0.2467 Episode_Reward/pair_body_intrusion: -0.2197 Curriculum/terrain_level_mean: +1.1606 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.3226 Episode_Termination/fell_over: 0.0968 Episode_Termination/course_exit: 0.5806 Episode_Termination/illegal_nonwheel_contact: 0.0065 ============================================================================== ============================================================================== Learning iteration 5200/50000 Total steps: 21,299,200 Steps per second: 7,634 Iteration time: 0.537s Collection time: 0.104s Learning time: 0.411s Mean reward: +0.0848 Mean episode length: 1289.6 Actor loss: -9.4317 Critic loss: +0.0033 World model loss: +1.5216 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8005 Episode_Reward/track_angular_velocity: +48.9274 Episode_Reward/upright: +25.0472 Episode_Reward/leg_posture: -3.0494 Episode_Reward/standstill: -6.1054 Episode_Reward/action_rate_l2: -0.3892 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0428 Episode_Reward/unnecessary_wheel_lift: -0.1435 Episode_Reward/rolling_leg_motion: -0.7886 Episode_Reward/obstacle_progress: +1.7298 Episode_Reward/wheel_body_intrusion: -0.2890 Episode_Reward/pair_body_intrusion: -0.1892 Curriculum/terrain_level_mean: +1.1711 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1030 Metrics/base_velocity/error_vel_yaw: 0.1316 Episode_Termination/time_out: 0.4178 Episode_Termination/fell_over: 0.1096 Episode_Termination/course_exit: 0.4658 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2265] Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 5250/50000 Total steps: 21,504,000 Steps per second: 7,633 Iteration time: 0.537s Collection time: 0.110s Learning time: 0.431s Mean reward: +0.0849 Mean episode length: 1352.0 Actor loss: -9.4143 Critic loss: +0.0029 World model loss: +1.4113 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0007 Episode_Reward/track_angular_velocity: +51.3104 Episode_Reward/upright: +26.2787 Episode_Reward/leg_posture: -3.2211 Episode_Reward/standstill: -6.3750 Episode_Reward/action_rate_l2: -0.4046 Episode_Reward/joint_torques_l2: -0.1542 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.1503 Episode_Reward/rolling_leg_motion: -0.8256 Episode_Reward/obstacle_progress: +1.8380 Episode_Reward/wheel_body_intrusion: -0.2898 Episode_Reward/pair_body_intrusion: -0.2316 Curriculum/terrain_level_mean: +1.1826 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1030 Metrics/base_velocity/error_vel_yaw: 0.1312 Episode_Termination/time_out: 0.4808 Episode_Termination/fell_over: 0.0513 Episode_Termination/course_exit: 0.4615 Episode_Termination/illegal_nonwheel_contact: 0.0064 ============================================================================== ============================================================================== Learning iteration 5300/50000 Total steps: 21,708,800 Steps per second: 7,631 Iteration time: 0.537s Collection time: 0.110s Learning time: 0.441s Mean reward: +0.0850 Mean episode length: 1336.7 Actor loss: -9.4121 Critic loss: +0.0039 World model loss: +1.4502 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4706 Episode_Reward/track_angular_velocity: +50.7701 Episode_Reward/upright: +25.9723 Episode_Reward/leg_posture: -3.1489 Episode_Reward/standstill: -6.1241 Episode_Reward/action_rate_l2: -0.4043 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1538 Episode_Reward/rolling_leg_motion: -0.8044 Episode_Reward/obstacle_progress: +1.7961 Episode_Reward/wheel_body_intrusion: -0.2886 Episode_Reward/pair_body_intrusion: -0.2454 Curriculum/terrain_level_mean: +1.1919 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1302 Episode_Termination/time_out: 0.4758 Episode_Termination/fell_over: 0.0484 Episode_Termination/course_exit: 0.4677 Episode_Termination/illegal_nonwheel_contact: 0.0081 ============================================================================== ============================================================================== Learning iteration 5350/50000 Total steps: 21,913,600 Steps per second: 7,630 Iteration time: 0.537s Collection time: 0.110s Learning time: 0.435s Mean reward: +0.0846 Mean episode length: 1258.2 Actor loss: -9.4242 Critic loss: +0.0045 World model loss: +1.4285 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.6326 Episode_Reward/track_angular_velocity: +47.7252 Episode_Reward/upright: +24.4358 Episode_Reward/leg_posture: -3.0413 Episode_Reward/standstill: -5.9827 Episode_Reward/action_rate_l2: -0.3872 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0442 Episode_Reward/unnecessary_wheel_lift: -0.1806 Episode_Reward/rolling_leg_motion: -0.7687 Episode_Reward/obstacle_progress: +1.6923 Episode_Reward/wheel_body_intrusion: -0.2768 Episode_Reward/pair_body_intrusion: -0.2219 Curriculum/terrain_level_mean: +1.2019 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1040 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.3493 Episode_Termination/fell_over: 0.1164 Episode_Termination/course_exit: 0.5274 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== ============================================================================== Learning iteration 5400/50000 Total steps: 22,118,400 Steps per second: 7,629 Iteration time: 0.537s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0850 Mean episode length: 1269.6 Actor loss: -9.4322 Critic loss: +0.0031 World model loss: +1.4730 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.1384 Episode_Reward/track_angular_velocity: +48.1870 Episode_Reward/upright: +24.6828 Episode_Reward/leg_posture: -2.9834 Episode_Reward/standstill: -5.9269 Episode_Reward/action_rate_l2: -0.3999 Episode_Reward/joint_torques_l2: -0.1435 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.1342 Episode_Reward/rolling_leg_motion: -0.7783 Episode_Reward/obstacle_progress: +1.7122 Episode_Reward/wheel_body_intrusion: -0.2478 Episode_Reward/pair_body_intrusion: -0.1879 Curriculum/terrain_level_mean: +1.2144 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1312 Episode_Termination/time_out: 0.3846 Episode_Termination/fell_over: 0.0833 Episode_Termination/course_exit: 0.5321 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_005400.mp4 ============================================================================== Learning iteration 5450/50000 Total steps: 22,323,200 Steps per second: 7,609 Iteration time: 0.538s Collection time: 0.110s Learning time: 0.431s Mean reward: +0.0850 Mean episode length: 1381.8 Actor loss: -9.4182 Critic loss: +0.0036 World model loss: +1.3859 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1282 Episode_Reward/track_angular_velocity: +52.4425 Episode_Reward/upright: +26.8807 Episode_Reward/leg_posture: -3.2770 Episode_Reward/standstill: -6.3574 Episode_Reward/action_rate_l2: -0.4519 Episode_Reward/joint_torques_l2: -0.1580 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.1354 Episode_Reward/rolling_leg_motion: -0.8476 Episode_Reward/obstacle_progress: +1.8504 Episode_Reward/wheel_body_intrusion: -0.2915 Episode_Reward/pair_body_intrusion: -0.2478 Curriculum/terrain_level_mean: +1.2263 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1021 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.5062 Episode_Termination/fell_over: 0.0375 Episode_Termination/course_exit: 0.4562 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 5500/50000 Total steps: 22,528,000 Steps per second: 7,609 Iteration time: 0.538s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0850 Mean episode length: 1339.0 Actor loss: -9.4301 Critic loss: +0.0105 World model loss: +1.4177 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6988 Episode_Reward/track_angular_velocity: +50.8357 Episode_Reward/upright: +26.0620 Episode_Reward/leg_posture: -3.1635 Episode_Reward/standstill: -6.3420 Episode_Reward/action_rate_l2: -0.4295 Episode_Reward/joint_torques_l2: -0.1495 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0288 Episode_Reward/unnecessary_wheel_lift: -0.1377 Episode_Reward/rolling_leg_motion: -0.8200 Episode_Reward/obstacle_progress: +1.7686 Episode_Reward/wheel_body_intrusion: -0.2532 Episode_Reward/pair_body_intrusion: -0.2268 Curriculum/terrain_level_mean: +1.2397 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1018 Metrics/base_velocity/error_vel_yaw: 0.1310 Episode_Termination/time_out: 0.4780 Episode_Termination/fell_over: 0.0566 Episode_Termination/course_exit: 0.4528 Episode_Termination/illegal_nonwheel_contact: 0.0126 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1166] ============================================================================== Learning iteration 5550/50000 Total steps: 22,732,800 Steps per second: 7,607 Iteration time: 0.538s Collection time: 0.110s Learning time: 0.443s Mean reward: +0.0850 Mean episode length: 1372.1 Actor loss: -9.4005 Critic loss: +0.0045 World model loss: +1.4284 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7251 Episode_Reward/track_angular_velocity: +52.0917 Episode_Reward/upright: +26.7176 Episode_Reward/leg_posture: -3.1722 Episode_Reward/standstill: -6.4440 Episode_Reward/action_rate_l2: -0.4381 Episode_Reward/joint_torques_l2: -0.1513 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0295 Episode_Reward/unnecessary_wheel_lift: -0.1387 Episode_Reward/rolling_leg_motion: -0.8327 Episode_Reward/obstacle_progress: +1.7783 Episode_Reward/wheel_body_intrusion: -0.2767 Episode_Reward/pair_body_intrusion: -0.2354 Curriculum/terrain_level_mean: +1.2510 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1023 Metrics/base_velocity/error_vel_yaw: 0.1308 Episode_Termination/time_out: 0.5515 Episode_Termination/fell_over: 0.0667 Episode_Termination/course_exit: 0.3818 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 5600/50000 Total steps: 22,937,600 Steps per second: 7,606 Iteration time: 0.539s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0848 Mean episode length: 1431.6 Actor loss: -9.4043 Critic loss: +0.0035 World model loss: +1.4075 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6771 Episode_Reward/track_angular_velocity: +54.2750 Episode_Reward/upright: +27.8880 Episode_Reward/leg_posture: -3.3615 Episode_Reward/standstill: -6.6983 Episode_Reward/action_rate_l2: -0.4629 Episode_Reward/joint_torques_l2: -0.1578 Episode_Reward/dof_pos_limits: -0.0019 Episode_Reward/terminated: -0.0056 Episode_Reward/unnecessary_wheel_lift: -0.1420 Episode_Reward/rolling_leg_motion: -0.8589 Episode_Reward/obstacle_progress: +1.8604 Episode_Reward/wheel_body_intrusion: -0.2889 Episode_Reward/pair_body_intrusion: -0.2760 Curriculum/terrain_level_mean: +1.2644 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.5978 Episode_Termination/fell_over: 0.0112 Episode_Termination/course_exit: 0.3911 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1997] ============================================================================== Learning iteration 5650/50000 Total steps: 23,142,400 Steps per second: 7,606 Iteration time: 0.539s Collection time: 0.110s Learning time: 0.432s Mean reward: +0.0842 Mean episode length: 1444.1 Actor loss: -9.4367 Critic loss: +0.0026 World model loss: +1.3570 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0331 Episode_Reward/track_angular_velocity: +54.7130 Episode_Reward/upright: +28.1198 Episode_Reward/leg_posture: -3.3835 Episode_Reward/standstill: -7.4268 Episode_Reward/action_rate_l2: -0.4925 Episode_Reward/joint_torques_l2: -0.1604 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0423 Episode_Reward/unnecessary_wheel_lift: -0.1620 Episode_Reward/rolling_leg_motion: -0.9032 Episode_Reward/obstacle_progress: +1.8179 Episode_Reward/wheel_body_intrusion: -0.2806 Episode_Reward/pair_body_intrusion: -0.2732 Curriculum/terrain_level_mean: +1.2749 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.6132 Episode_Termination/fell_over: 0.0708 Episode_Termination/course_exit: 0.3160 Episode_Termination/illegal_nonwheel_contact: 0.0047 ============================================================================== ============================================================================== Learning iteration 5700/50000 Total steps: 23,347,200 Steps per second: 7,631 Iteration time: 0.537s Collection time: 0.086s Learning time: 0.248s Mean reward: +0.0846 Mean episode length: 1454.2 Actor loss: -9.4223 Critic loss: +0.0026 World model loss: +1.3427 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.4793 Episode_Reward/track_angular_velocity: +55.0361 Episode_Reward/upright: +28.3227 Episode_Reward/leg_posture: -3.3160 Episode_Reward/standstill: -6.9917 Episode_Reward/action_rate_l2: -0.5174 Episode_Reward/joint_torques_l2: -0.1592 Episode_Reward/dof_pos_limits: -0.0001 Episode_Reward/terminated: -0.0398 Episode_Reward/unnecessary_wheel_lift: -0.1864 Episode_Reward/rolling_leg_motion: -0.9163 Episode_Reward/obstacle_progress: +1.8283 Episode_Reward/wheel_body_intrusion: -0.2794 Episode_Reward/pair_body_intrusion: -0.2509 Curriculum/terrain_level_mean: +1.2856 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1026 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.6398 Episode_Termination/fell_over: 0.0551 Episode_Termination/course_exit: 0.3008 Episode_Termination/illegal_nonwheel_contact: 0.0085 ============================================================================== ============================================================================== Learning iteration 5750/50000 Total steps: 23,552,000 Steps per second: 7,666 Iteration time: 0.534s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0847 Mean episode length: 1473.0 Actor loss: -9.4256 Critic loss: +0.0027 World model loss: +1.3875 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.1404 Episode_Reward/track_angular_velocity: +55.7540 Episode_Reward/upright: +28.7124 Episode_Reward/leg_posture: -3.3525 Episode_Reward/standstill: -6.9695 Episode_Reward/action_rate_l2: -0.5410 Episode_Reward/joint_torques_l2: -0.1602 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0345 Episode_Reward/unnecessary_wheel_lift: -0.1595 Episode_Reward/rolling_leg_motion: -0.9419 Episode_Reward/obstacle_progress: +1.8397 Episode_Reward/wheel_body_intrusion: -0.3382 Episode_Reward/pair_body_intrusion: -0.2407 Curriculum/terrain_level_mean: +1.2961 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.6638 Episode_Termination/fell_over: 0.0393 Episode_Termination/course_exit: 0.2838 Episode_Termination/illegal_nonwheel_contact: 0.0131 ============================================================================== ============================================================================== Learning iteration 5800/50000 Total steps: 23,756,800 Steps per second: 7,700 Iteration time: 0.532s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0847 Mean episode length: 1444.4 Actor loss: -9.4263 Critic loss: +0.0035 World model loss: +1.4471 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1921 Episode_Reward/track_angular_velocity: +54.6168 Episode_Reward/upright: +28.1874 Episode_Reward/leg_posture: -3.2679 Episode_Reward/standstill: -6.7822 Episode_Reward/action_rate_l2: -0.5580 Episode_Reward/joint_torques_l2: -0.1586 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0310 Episode_Reward/unnecessary_wheel_lift: -0.1449 Episode_Reward/rolling_leg_motion: -0.9335 Episode_Reward/obstacle_progress: +1.8245 Episode_Reward/wheel_body_intrusion: -0.3287 Episode_Reward/pair_body_intrusion: -0.2859 Curriculum/terrain_level_mean: +1.3076 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1026 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.6162 Episode_Termination/fell_over: 0.0556 Episode_Termination/course_exit: 0.3283 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 5850/50000 Total steps: 23,961,600 Steps per second: 7,734 Iteration time: 0.530s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0848 Mean episode length: 1396.0 Actor loss: -9.4347 Critic loss: +0.0026 World model loss: +1.2956 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4675 Episode_Reward/track_angular_velocity: +52.8182 Episode_Reward/upright: +27.2453 Episode_Reward/leg_posture: -3.1720 Episode_Reward/standstill: -6.4447 Episode_Reward/action_rate_l2: -0.5380 Episode_Reward/joint_torques_l2: -0.1524 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.1304 Episode_Reward/rolling_leg_motion: -0.8944 Episode_Reward/obstacle_progress: +1.7900 Episode_Reward/wheel_body_intrusion: -0.3130 Episode_Reward/pair_body_intrusion: -0.2383 Curriculum/terrain_level_mean: +1.3191 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5663 Episode_Termination/fell_over: 0.0361 Episode_Termination/course_exit: 0.3855 Episode_Termination/illegal_nonwheel_contact: 0.0120 ============================================================================== ============================================================================== Learning iteration 5900/50000 Total steps: 24,166,400 Steps per second: 7,768 Iteration time: 0.527s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0846 Mean episode length: 1370.1 Actor loss: -9.4130 Critic loss: +0.0025 World model loss: +1.3728 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5817 Episode_Reward/track_angular_velocity: +51.8853 Episode_Reward/upright: +26.7433 Episode_Reward/leg_posture: -3.1098 Episode_Reward/standstill: -6.6427 Episode_Reward/action_rate_l2: -0.5265 Episode_Reward/joint_torques_l2: -0.1534 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0321 Episode_Reward/unnecessary_wheel_lift: -0.1579 Episode_Reward/rolling_leg_motion: -0.8661 Episode_Reward/obstacle_progress: +1.7421 Episode_Reward/wheel_body_intrusion: -0.3058 Episode_Reward/pair_body_intrusion: -0.2438 Curriculum/terrain_level_mean: +1.3289 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1026 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5521 Episode_Termination/fell_over: 0.0613 Episode_Termination/course_exit: 0.3742 Episode_Termination/illegal_nonwheel_contact: 0.0184 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2200] ============================================================================== Learning iteration 5950/50000 Total steps: 24,371,200 Steps per second: 7,801 Iteration time: 0.525s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0846 Mean episode length: 1284.4 Actor loss: -9.4271 Critic loss: +0.0035 World model loss: +1.3641 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.6234 Episode_Reward/track_angular_velocity: +48.6299 Episode_Reward/upright: +25.0432 Episode_Reward/leg_posture: -2.9207 Episode_Reward/standstill: -6.1843 Episode_Reward/action_rate_l2: -0.4956 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0452 Episode_Reward/unnecessary_wheel_lift: -0.1495 Episode_Reward/rolling_leg_motion: -0.8134 Episode_Reward/obstacle_progress: +1.6016 Episode_Reward/wheel_body_intrusion: -0.2753 Episode_Reward/pair_body_intrusion: -0.2013 Curriculum/terrain_level_mean: +1.3389 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1024 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4400 Episode_Termination/fell_over: 0.0971 Episode_Termination/course_exit: 0.4571 Episode_Termination/illegal_nonwheel_contact: 0.0114 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2138] ============================================================================== Learning iteration 6000/50000 Total steps: 24,576,000 Steps per second: 7,834 Iteration time: 0.523s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0843 Mean episode length: 1302.0 Actor loss: -9.4111 Critic loss: +0.0024 World model loss: +1.3713 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.1722 Episode_Reward/track_angular_velocity: +49.2876 Episode_Reward/upright: +25.3791 Episode_Reward/leg_posture: -3.0207 Episode_Reward/standstill: -6.4595 Episode_Reward/action_rate_l2: -0.5045 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0432 Episode_Reward/unnecessary_wheel_lift: -0.1673 Episode_Reward/rolling_leg_motion: -0.8329 Episode_Reward/obstacle_progress: +1.5889 Episode_Reward/wheel_body_intrusion: -0.3099 Episode_Reward/pair_body_intrusion: -0.2306 Curriculum/terrain_level_mean: +1.3481 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1029 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4459 Episode_Termination/fell_over: 0.1019 Episode_Termination/course_exit: 0.4459 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 6050/50000 Total steps: 24,780,800 Steps per second: 7,866 Iteration time: 0.521s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0845 Mean episode length: 1298.5 Actor loss: -9.4172 Critic loss: +0.0048 World model loss: +1.3496 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.0992 Episode_Reward/track_angular_velocity: +49.0972 Episode_Reward/upright: +25.3203 Episode_Reward/leg_posture: -2.9855 Episode_Reward/standstill: -6.3596 Episode_Reward/action_rate_l2: -0.5095 Episode_Reward/joint_torques_l2: -0.1472 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0304 Episode_Reward/unnecessary_wheel_lift: -0.1358 Episode_Reward/rolling_leg_motion: -0.8009 Episode_Reward/obstacle_progress: +1.6631 Episode_Reward/wheel_body_intrusion: -0.3076 Episode_Reward/pair_body_intrusion: -0.2446 Curriculum/terrain_level_mean: +1.3574 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1023 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4609 Episode_Termination/fell_over: 0.0938 Episode_Termination/course_exit: 0.4453 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 6100/50000 Total steps: 24,985,600 Steps per second: 7,898 Iteration time: 0.519s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0841 Mean episode length: 1342.0 Actor loss: -9.3806 Critic loss: +0.0066 World model loss: +1.3984 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4601 Episode_Reward/track_angular_velocity: +50.7312 Episode_Reward/upright: +26.1495 Episode_Reward/leg_posture: -3.0619 Episode_Reward/standstill: -6.9605 Episode_Reward/action_rate_l2: -0.5365 Episode_Reward/joint_torques_l2: -0.1543 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0341 Episode_Reward/unnecessary_wheel_lift: -0.1169 Episode_Reward/rolling_leg_motion: -0.8453 Episode_Reward/obstacle_progress: +1.7427 Episode_Reward/wheel_body_intrusion: -0.3258 Episode_Reward/pair_body_intrusion: -0.2289 Curriculum/terrain_level_mean: +1.3650 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1036 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4583 Episode_Termination/fell_over: 0.0903 Episode_Termination/course_exit: 0.4514 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[661] ============================================================================== Learning iteration 6150/50000 Total steps: 25,190,400 Steps per second: 7,931 Iteration time: 0.516s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0842 Mean episode length: 1239.4 Actor loss: -9.4032 Critic loss: +0.0055 World model loss: +1.3149 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.9730 Episode_Reward/track_angular_velocity: +46.8822 Episode_Reward/upright: +24.1542 Episode_Reward/leg_posture: -2.8226 Episode_Reward/standstill: -6.3618 Episode_Reward/action_rate_l2: -0.5007 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0194 Episode_Reward/unnecessary_wheel_lift: -0.1110 Episode_Reward/rolling_leg_motion: -0.7706 Episode_Reward/obstacle_progress: +1.6069 Episode_Reward/wheel_body_intrusion: -0.3079 Episode_Reward/pair_body_intrusion: -0.2211 Curriculum/terrain_level_mean: +1.3752 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1030 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4211 Episode_Termination/fell_over: 0.0702 Episode_Termination/course_exit: 0.5088 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 6200/50000 Total steps: 25,395,200 Steps per second: 7,963 Iteration time: 0.514s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0839 Mean episode length: 1315.4 Actor loss: -9.4183 Critic loss: +0.0027 World model loss: +1.3390 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.4016 Episode_Reward/track_angular_velocity: +49.7918 Episode_Reward/upright: +25.6189 Episode_Reward/leg_posture: -3.0053 Episode_Reward/standstill: -6.8038 Episode_Reward/action_rate_l2: -0.5329 Episode_Reward/joint_torques_l2: -0.1533 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0283 Episode_Reward/unnecessary_wheel_lift: -0.1501 Episode_Reward/rolling_leg_motion: -0.8178 Episode_Reward/obstacle_progress: +1.6681 Episode_Reward/wheel_body_intrusion: -0.3328 Episode_Reward/pair_body_intrusion: -0.2731 Curriculum/terrain_level_mean: +1.3835 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4526 Episode_Termination/fell_over: 0.0730 Episode_Termination/course_exit: 0.4672 Episode_Termination/illegal_nonwheel_contact: 0.0146 ============================================================================== ============================================================================== Learning iteration 6250/50000 Total steps: 25,600,000 Steps per second: 7,994 Iteration time: 0.512s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0841 Mean episode length: 1240.9 Actor loss: -9.3968 Critic loss: +0.0030 World model loss: +1.3735 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.9436 Episode_Reward/track_angular_velocity: +46.9299 Episode_Reward/upright: +24.1566 Episode_Reward/leg_posture: -2.7861 Episode_Reward/standstill: -6.3700 Episode_Reward/action_rate_l2: -0.5169 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.1339 Episode_Reward/rolling_leg_motion: -0.7638 Episode_Reward/obstacle_progress: +1.6630 Episode_Reward/wheel_body_intrusion: -0.3085 Episode_Reward/pair_body_intrusion: -0.2397 Curriculum/terrain_level_mean: +1.3938 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1039 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4370 Episode_Termination/fell_over: 0.0741 Episode_Termination/course_exit: 0.4815 Episode_Termination/illegal_nonwheel_contact: 0.0148 ============================================================================== ============================================================================== Learning iteration 6300/50000 Total steps: 25,804,800 Steps per second: 8,026 Iteration time: 0.510s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0839 Mean episode length: 1300.6 Actor loss: -9.3951 Critic loss: +0.0045 World model loss: +1.3106 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8868 Episode_Reward/track_angular_velocity: +49.1480 Episode_Reward/upright: +25.3027 Episode_Reward/leg_posture: -2.9455 Episode_Reward/standstill: -6.7819 Episode_Reward/action_rate_l2: -0.5542 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0254 Episode_Reward/unnecessary_wheel_lift: -0.1282 Episode_Reward/rolling_leg_motion: -0.8217 Episode_Reward/obstacle_progress: +1.7118 Episode_Reward/wheel_body_intrusion: -0.3209 Episode_Reward/pair_body_intrusion: -0.2443 Curriculum/terrain_level_mean: +1.4043 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0625 Episode_Termination/course_exit: 0.4861 Episode_Termination/illegal_nonwheel_contact: 0.0069 ============================================================================== ============================================================================== Learning iteration 6350/50000 Total steps: 26,009,600 Steps per second: 8,056 Iteration time: 0.508s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0839 Mean episode length: 1291.4 Actor loss: -9.3851 Critic loss: +0.0065 World model loss: +1.3168 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.5701 Episode_Reward/track_angular_velocity: +48.8661 Episode_Reward/upright: +25.1136 Episode_Reward/leg_posture: -2.9393 Episode_Reward/standstill: -6.7481 Episode_Reward/action_rate_l2: -0.5387 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0353 Episode_Reward/unnecessary_wheel_lift: -0.1094 Episode_Reward/rolling_leg_motion: -0.7975 Episode_Reward/obstacle_progress: +1.6740 Episode_Reward/wheel_body_intrusion: -0.3061 Episode_Reward/pair_body_intrusion: -0.2397 Curriculum/terrain_level_mean: +1.4124 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4252 Episode_Termination/fell_over: 0.0945 Episode_Termination/course_exit: 0.4646 Episode_Termination/illegal_nonwheel_contact: 0.0315 ============================================================================== ============================================================================== Learning iteration 6400/50000 Total steps: 26,214,400 Steps per second: 8,086 Iteration time: 0.507s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0836 Mean episode length: 1242.9 Actor loss: -9.4173 Critic loss: +0.0057 World model loss: +1.2763 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.8263 Episode_Reward/track_angular_velocity: +47.0519 Episode_Reward/upright: +24.1405 Episode_Reward/leg_posture: -2.8584 Episode_Reward/standstill: -6.7784 Episode_Reward/action_rate_l2: -0.5192 Episode_Reward/joint_torques_l2: -0.1465 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0364 Episode_Reward/unnecessary_wheel_lift: -0.1350 Episode_Reward/rolling_leg_motion: -0.7627 Episode_Reward/obstacle_progress: +1.6338 Episode_Reward/wheel_body_intrusion: -0.2961 Episode_Reward/pair_body_intrusion: -0.2769 Curriculum/terrain_level_mean: +1.4224 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1053 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.3836 Episode_Termination/fell_over: 0.0959 Episode_Termination/course_exit: 0.5137 Episode_Termination/illegal_nonwheel_contact: 0.0137 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[976] ============================================================================== Learning iteration 6450/50000 Total steps: 26,419,200 Steps per second: 8,117 Iteration time: 0.505s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0835 Mean episode length: 1345.3 Actor loss: -9.4023 Critic loss: +0.0045 World model loss: +1.2962 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2980 Episode_Reward/track_angular_velocity: +50.8692 Episode_Reward/upright: +26.1288 Episode_Reward/leg_posture: -3.0455 Episode_Reward/standstill: -7.3709 Episode_Reward/action_rate_l2: -0.5849 Episode_Reward/joint_torques_l2: -0.1596 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0447 Episode_Reward/unnecessary_wheel_lift: -0.1639 Episode_Reward/rolling_leg_motion: -0.8374 Episode_Reward/obstacle_progress: +1.7522 Episode_Reward/wheel_body_intrusion: -0.3061 Episode_Reward/pair_body_intrusion: -0.2351 Curriculum/terrain_level_mean: +1.4302 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1060 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4863 Episode_Termination/fell_over: 0.1027 Episode_Termination/course_exit: 0.3973 Episode_Termination/illegal_nonwheel_contact: 0.0137 ============================================================================== ============================================================================== Learning iteration 6500/50000 Total steps: 26,624,000 Steps per second: 8,147 Iteration time: 0.503s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0838 Mean episode length: 1201.2 Actor loss: -9.3878 Critic loss: +0.0087 World model loss: +1.2695 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.4287 Episode_Reward/track_angular_velocity: +45.3758 Episode_Reward/upright: +23.3201 Episode_Reward/leg_posture: -2.7132 Episode_Reward/standstill: -6.2583 Episode_Reward/action_rate_l2: -0.5565 Episode_Reward/joint_torques_l2: -0.1443 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0282 Episode_Reward/unnecessary_wheel_lift: -0.1147 Episode_Reward/rolling_leg_motion: -0.7415 Episode_Reward/obstacle_progress: +1.6486 Episode_Reward/wheel_body_intrusion: -0.2858 Episode_Reward/pair_body_intrusion: -0.2282 Curriculum/terrain_level_mean: +1.4387 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1051 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.3500 Episode_Termination/fell_over: 0.0917 Episode_Termination/course_exit: 0.5500 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== ============================================================================== Learning iteration 6550/50000 Total steps: 26,828,800 Steps per second: 8,177 Iteration time: 0.501s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0841 Mean episode length: 1340.0 Actor loss: -9.3873 Critic loss: +0.0106 World model loss: +1.2722 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3224 Episode_Reward/track_angular_velocity: +50.6677 Episode_Reward/upright: +26.0181 Episode_Reward/leg_posture: -2.9616 Episode_Reward/standstill: -6.9144 Episode_Reward/action_rate_l2: -0.6181 Episode_Reward/joint_torques_l2: -0.1606 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0445 Episode_Reward/unnecessary_wheel_lift: -0.1686 Episode_Reward/rolling_leg_motion: -0.8232 Episode_Reward/obstacle_progress: +1.9343 Episode_Reward/wheel_body_intrusion: -0.2868 Episode_Reward/pair_body_intrusion: -0.2049 Curriculum/terrain_level_mean: +1.4448 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4571 Episode_Termination/fell_over: 0.1214 Episode_Termination/course_exit: 0.4214 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 6600/50000 Total steps: 27,033,600 Steps per second: 8,207 Iteration time: 0.499s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0844 Mean episode length: 1309.6 Actor loss: -9.4018 Critic loss: +0.0033 World model loss: +1.2731 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.2728 Episode_Reward/track_angular_velocity: +49.4021 Episode_Reward/upright: +25.4137 Episode_Reward/leg_posture: -2.9106 Episode_Reward/standstill: -6.2457 Episode_Reward/action_rate_l2: -0.5893 Episode_Reward/joint_torques_l2: -0.1585 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0307 Episode_Reward/unnecessary_wheel_lift: -0.1323 Episode_Reward/rolling_leg_motion: -0.7970 Episode_Reward/obstacle_progress: +1.9104 Episode_Reward/wheel_body_intrusion: -0.3043 Episode_Reward/pair_body_intrusion: -0.2556 Curriculum/terrain_level_mean: +1.4541 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1045 Metrics/base_velocity/error_vel_yaw: 0.1392 Episode_Termination/time_out: 0.4483 Episode_Termination/fell_over: 0.0759 Episode_Termination/course_exit: 0.4690 Episode_Termination/illegal_nonwheel_contact: 0.0207 ============================================================================== ============================================================================== Learning iteration 6650/50000 Total steps: 27,238,400 Steps per second: 8,236 Iteration time: 0.497s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0845 Mean episode length: 1344.0 Actor loss: -9.3992 Critic loss: +0.0074 World model loss: +1.3024 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5517 Episode_Reward/track_angular_velocity: +50.8197 Episode_Reward/upright: +26.0991 Episode_Reward/leg_posture: -2.9932 Episode_Reward/standstill: -6.5527 Episode_Reward/action_rate_l2: -0.6099 Episode_Reward/joint_torques_l2: -0.1644 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0236 Episode_Reward/unnecessary_wheel_lift: -0.1171 Episode_Reward/rolling_leg_motion: -0.8203 Episode_Reward/obstacle_progress: +1.9803 Episode_Reward/wheel_body_intrusion: -0.3196 Episode_Reward/pair_body_intrusion: -0.2409 Curriculum/terrain_level_mean: +1.4622 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4667 Episode_Termination/fell_over: 0.0593 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0222 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2262] ============================================================================== Learning iteration 6700/50000 Total steps: 27,443,200 Steps per second: 8,266 Iteration time: 0.496s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1332.4 Actor loss: -9.4259 Critic loss: +0.0048 World model loss: +1.3096 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2268 Episode_Reward/track_angular_velocity: +50.3434 Episode_Reward/upright: +25.8621 Episode_Reward/leg_posture: -3.0093 Episode_Reward/standstill: -6.4709 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1632 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0182 Episode_Reward/unnecessary_wheel_lift: -0.1157 Episode_Reward/rolling_leg_motion: -0.8042 Episode_Reward/obstacle_progress: +1.9884 Episode_Reward/wheel_body_intrusion: -0.3131 Episode_Reward/pair_body_intrusion: -0.2057 Curriculum/terrain_level_mean: +1.4690 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5043 Episode_Termination/fell_over: 0.0598 Episode_Termination/course_exit: 0.4359 Episode_Termination/illegal_nonwheel_contact: 0.0171 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1866] ============================================================================== Learning iteration 6750/50000 Total steps: 27,648,000 Steps per second: 8,295 Iteration time: 0.494s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0846 Mean episode length: 1272.3 Actor loss: -9.4179 Critic loss: +0.0032 World model loss: +1.2711 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.0370 Episode_Reward/track_angular_velocity: +48.0494 Episode_Reward/upright: +24.7019 Episode_Reward/leg_posture: -2.8523 Episode_Reward/standstill: -6.1472 Episode_Reward/action_rate_l2: -0.5315 Episode_Reward/joint_torques_l2: -0.1551 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0298 Episode_Reward/unnecessary_wheel_lift: -0.1193 Episode_Reward/rolling_leg_motion: -0.7508 Episode_Reward/obstacle_progress: +1.8941 Episode_Reward/wheel_body_intrusion: -0.2812 Episode_Reward/pair_body_intrusion: -0.2013 Curriculum/terrain_level_mean: +1.4761 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1375 Episode_Termination/time_out: 0.4035 Episode_Termination/fell_over: 0.1053 Episode_Termination/course_exit: 0.4912 Episode_Termination/illegal_nonwheel_contact: 0.0088 ============================================================================== ============================================================================== Learning iteration 6800/50000 Total steps: 27,852,800 Steps per second: 8,324 Iteration time: 0.492s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1238.3 Actor loss: -9.3973 Critic loss: +0.0089 World model loss: +1.3099 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.8543 Episode_Reward/track_angular_velocity: +46.7912 Episode_Reward/upright: +24.0425 Episode_Reward/leg_posture: -2.8206 Episode_Reward/standstill: -5.7614 Episode_Reward/action_rate_l2: -0.5193 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0314 Episode_Reward/unnecessary_wheel_lift: -0.1306 Episode_Reward/rolling_leg_motion: -0.7445 Episode_Reward/obstacle_progress: +1.8373 Episode_Reward/wheel_body_intrusion: -0.3139 Episode_Reward/pair_body_intrusion: -0.2395 Curriculum/terrain_level_mean: +1.4880 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.3600 Episode_Termination/fell_over: 0.0800 Episode_Termination/course_exit: 0.5533 Episode_Termination/illegal_nonwheel_contact: 0.0133 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[839] ============================================================================== Learning iteration 6850/50000 Total steps: 28,057,600 Steps per second: 8,353 Iteration time: 0.490s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0848 Mean episode length: 1249.4 Actor loss: -9.4059 Critic loss: +0.0122 World model loss: +1.2676 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.3127 Episode_Reward/track_angular_velocity: +47.2193 Episode_Reward/upright: +24.2445 Episode_Reward/leg_posture: -2.8602 Episode_Reward/standstill: -5.7014 Episode_Reward/action_rate_l2: -0.5286 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0317 Episode_Reward/unnecessary_wheel_lift: -0.1176 Episode_Reward/rolling_leg_motion: -0.7479 Episode_Reward/obstacle_progress: +1.8606 Episode_Reward/wheel_body_intrusion: -0.3031 Episode_Reward/pair_body_intrusion: -0.2132 Curriculum/terrain_level_mean: +1.4978 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1039 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4167 Episode_Termination/fell_over: 0.0833 Episode_Termination/course_exit: 0.4931 Episode_Termination/illegal_nonwheel_contact: 0.0139 ============================================================================== ============================================================================== Learning iteration 6900/50000 Total steps: 28,262,400 Steps per second: 8,381 Iteration time: 0.489s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1335.8 Actor loss: -9.4023 Critic loss: +0.0064 World model loss: +1.2630 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2608 Episode_Reward/track_angular_velocity: +50.5514 Episode_Reward/upright: +25.9491 Episode_Reward/leg_posture: -3.0205 Episode_Reward/standstill: -6.2421 Episode_Reward/action_rate_l2: -0.5626 Episode_Reward/joint_torques_l2: -0.1598 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.1448 Episode_Reward/rolling_leg_motion: -0.7905 Episode_Reward/obstacle_progress: +1.9474 Episode_Reward/wheel_body_intrusion: -0.3324 Episode_Reward/pair_body_intrusion: -0.2609 Curriculum/terrain_level_mean: +1.5059 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1038 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4671 Episode_Termination/fell_over: 0.0658 Episode_Termination/course_exit: 0.4539 Episode_Termination/illegal_nonwheel_contact: 0.0329 ============================================================================== ============================================================================== Learning iteration 6950/50000 Total steps: 28,467,200 Steps per second: 8,409 Iteration time: 0.487s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0848 Mean episode length: 1349.8 Actor loss: -9.4363 Critic loss: +0.0036 World model loss: +1.2549 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8984 Episode_Reward/track_angular_velocity: +51.0888 Episode_Reward/upright: +26.2431 Episode_Reward/leg_posture: -3.0301 Episode_Reward/standstill: -6.5118 Episode_Reward/action_rate_l2: -0.5721 Episode_Reward/joint_torques_l2: -0.1613 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.1260 Episode_Reward/rolling_leg_motion: -0.8047 Episode_Reward/obstacle_progress: +1.9916 Episode_Reward/wheel_body_intrusion: -0.3185 Episode_Reward/pair_body_intrusion: -0.2424 Curriculum/terrain_level_mean: +1.5151 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4357 Episode_Termination/fell_over: 0.0500 Episode_Termination/course_exit: 0.5143 Episode_Termination/illegal_nonwheel_contact: 0.0143 ============================================================================== ============================================================================== Learning iteration 7000/50000 Total steps: 28,672,000 Steps per second: 8,437 Iteration time: 0.485s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0851 Mean episode length: 1328.4 Actor loss: -9.4232 Critic loss: +0.0050 World model loss: +1.2726 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1909 Episode_Reward/track_angular_velocity: +50.2962 Episode_Reward/upright: +25.8341 Episode_Reward/leg_posture: -2.9556 Episode_Reward/standstill: -6.0898 Episode_Reward/action_rate_l2: -0.5601 Episode_Reward/joint_torques_l2: -0.1584 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0311 Episode_Reward/unnecessary_wheel_lift: -0.1201 Episode_Reward/rolling_leg_motion: -0.7999 Episode_Reward/obstacle_progress: +1.9928 Episode_Reward/wheel_body_intrusion: -0.3074 Episode_Reward/pair_body_intrusion: -0.2294 Curriculum/terrain_level_mean: +1.5269 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5217 Episode_Termination/fell_over: 0.0745 Episode_Termination/course_exit: 0.4037 Episode_Termination/illegal_nonwheel_contact: 0.0311 ============================================================================== ============================================================================== Learning iteration 7050/50000 Total steps: 28,876,800 Steps per second: 8,464 Iteration time: 0.484s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0852 Mean episode length: 1336.9 Actor loss: -9.4167 Critic loss: +0.0051 World model loss: +1.2110 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4929 Episode_Reward/track_angular_velocity: +50.6016 Episode_Reward/upright: +26.0002 Episode_Reward/leg_posture: -2.9746 Episode_Reward/standstill: -6.1266 Episode_Reward/action_rate_l2: -0.5631 Episode_Reward/joint_torques_l2: -0.1594 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.1245 Episode_Reward/rolling_leg_motion: -0.8143 Episode_Reward/obstacle_progress: +2.0636 Episode_Reward/wheel_body_intrusion: -0.3239 Episode_Reward/pair_body_intrusion: -0.2022 Curriculum/terrain_level_mean: +1.5347 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4841 Episode_Termination/fell_over: 0.0635 Episode_Termination/course_exit: 0.4524 Episode_Termination/illegal_nonwheel_contact: 0.0079 ============================================================================== ============================================================================== Learning iteration 7100/50000 Total steps: 29,081,600 Steps per second: 8,492 Iteration time: 0.482s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0853 Mean episode length: 1315.3 Actor loss: -9.3748 Critic loss: +0.0145 World model loss: +1.2051 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7079 Episode_Reward/track_angular_velocity: +49.7863 Episode_Reward/upright: +25.5824 Episode_Reward/leg_posture: -2.9245 Episode_Reward/standstill: -5.8115 Episode_Reward/action_rate_l2: -0.5483 Episode_Reward/joint_torques_l2: -0.1562 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.1164 Episode_Reward/rolling_leg_motion: -0.7761 Episode_Reward/obstacle_progress: +2.0583 Episode_Reward/wheel_body_intrusion: -0.3274 Episode_Reward/pair_body_intrusion: -0.2607 Curriculum/terrain_level_mean: +1.5449 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4681 Episode_Termination/fell_over: 0.0638 Episode_Termination/course_exit: 0.4610 Episode_Termination/illegal_nonwheel_contact: 0.0213 ============================================================================== ============================================================================== Learning iteration 7150/50000 Total steps: 29,286,400 Steps per second: 8,519 Iteration time: 0.481s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0848 Mean episode length: 1282.3 Actor loss: -9.4164 Critic loss: +0.0034 World model loss: +1.1969 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.5000 Episode_Reward/track_angular_velocity: +48.5033 Episode_Reward/upright: +24.9526 Episode_Reward/leg_posture: -2.8420 Episode_Reward/standstill: -6.1919 Episode_Reward/action_rate_l2: -0.5370 Episode_Reward/joint_torques_l2: -0.1523 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0301 Episode_Reward/unnecessary_wheel_lift: -0.1543 Episode_Reward/rolling_leg_motion: -0.7801 Episode_Reward/obstacle_progress: +2.0227 Episode_Reward/wheel_body_intrusion: -0.3225 Episode_Reward/pair_body_intrusion: -0.2483 Curriculum/terrain_level_mean: +1.5544 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1034 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4157 Episode_Termination/fell_over: 0.0602 Episode_Termination/course_exit: 0.5120 Episode_Termination/illegal_nonwheel_contact: 0.0361 ============================================================================== ============================================================================== Learning iteration 7200/50000 Total steps: 29,491,200 Steps per second: 8,546 Iteration time: 0.479s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0850 Mean episode length: 1375.0 Actor loss: -9.4092 Critic loss: +0.0107 World model loss: +1.2478 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8912 Episode_Reward/track_angular_velocity: +52.0261 Episode_Reward/upright: +26.7561 Episode_Reward/leg_posture: -3.0522 Episode_Reward/standstill: -6.7785 Episode_Reward/action_rate_l2: -0.5581 Episode_Reward/joint_torques_l2: -0.1643 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.1309 Episode_Reward/rolling_leg_motion: -0.8144 Episode_Reward/obstacle_progress: +2.2295 Episode_Reward/wheel_body_intrusion: -0.3275 Episode_Reward/pair_body_intrusion: -0.2276 Curriculum/terrain_level_mean: +1.5613 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1022 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5106 Episode_Termination/fell_over: 0.0709 Episode_Termination/course_exit: 0.4043 Episode_Termination/illegal_nonwheel_contact: 0.0355 ============================================================================== ============================================================================== Learning iteration 7250/50000 Total steps: 29,696,000 Steps per second: 8,573 Iteration time: 0.478s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1408.1 Actor loss: -9.4162 Critic loss: +0.0040 World model loss: +1.2293 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9587 Episode_Reward/track_angular_velocity: +53.1918 Episode_Reward/upright: +27.3889 Episode_Reward/leg_posture: -3.0991 Episode_Reward/standstill: -7.0167 Episode_Reward/action_rate_l2: -0.5838 Episode_Reward/joint_torques_l2: -0.1684 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0275 Episode_Reward/unnecessary_wheel_lift: -0.1322 Episode_Reward/rolling_leg_motion: -0.8362 Episode_Reward/obstacle_progress: +2.2936 Episode_Reward/wheel_body_intrusion: -0.3467 Episode_Reward/pair_body_intrusion: -0.3025 Curriculum/terrain_level_mean: +1.5703 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1026 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5030 Episode_Termination/fell_over: 0.0473 Episode_Termination/course_exit: 0.4379 Episode_Termination/illegal_nonwheel_contact: 0.0178 ============================================================================== ============================================================================== Learning iteration 7300/50000 Total steps: 29,900,800 Steps per second: 8,600 Iteration time: 0.476s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1382.6 Actor loss: -9.3951 Critic loss: +0.0034 World model loss: +1.2529 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0180 Episode_Reward/track_angular_velocity: +52.2670 Episode_Reward/upright: +26.8818 Episode_Reward/leg_posture: -3.0834 Episode_Reward/standstill: -6.9541 Episode_Reward/action_rate_l2: -0.5669 Episode_Reward/joint_torques_l2: -0.1663 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0324 Episode_Reward/unnecessary_wheel_lift: -0.1779 Episode_Reward/rolling_leg_motion: -0.8373 Episode_Reward/obstacle_progress: +2.1950 Episode_Reward/wheel_body_intrusion: -0.3673 Episode_Reward/pair_body_intrusion: -0.2763 Curriculum/terrain_level_mean: +1.5796 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.5200 Episode_Termination/fell_over: 0.0600 Episode_Termination/course_exit: 0.4200 Episode_Termination/illegal_nonwheel_contact: 0.0100 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2156] RECOVERY reset 1 invalid environments; first IDs=[2359] RECOVERY reset 1 invalid environments; first IDs=[1704] ============================================================================== Learning iteration 7350/50000 Total steps: 30,105,600 Steps per second: 8,626 Iteration time: 0.475s Collection time: 0.077s Learning time: 0.184s Mean reward: +0.0850 Mean episode length: 1372.8 Actor loss: -9.4259 Critic loss: +0.0052 World model loss: +1.2124 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8635 Episode_Reward/track_angular_velocity: +51.9394 Episode_Reward/upright: +26.7027 Episode_Reward/leg_posture: -3.0421 Episode_Reward/standstill: -6.5484 Episode_Reward/action_rate_l2: -0.5892 Episode_Reward/joint_torques_l2: -0.1646 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.1610 Episode_Reward/rolling_leg_motion: -0.8221 Episode_Reward/obstacle_progress: +2.2220 Episode_Reward/wheel_body_intrusion: -0.3547 Episode_Reward/pair_body_intrusion: -0.2748 Curriculum/terrain_level_mean: +1.5889 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1016 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.5444 Episode_Termination/fell_over: 0.0667 Episode_Termination/course_exit: 0.3889 Episode_Termination/illegal_nonwheel_contact: 0.0056 ============================================================================== ============================================================================== Learning iteration 7400/50000 Total steps: 30,310,400 Steps per second: 8,653 Iteration time: 0.473s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0850 Mean episode length: 1439.3 Actor loss: -9.4271 Critic loss: +0.0032 World model loss: +1.2461 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1201 Episode_Reward/track_angular_velocity: +54.3581 Episode_Reward/upright: +28.0038 Episode_Reward/leg_posture: -3.1420 Episode_Reward/standstill: -6.8968 Episode_Reward/action_rate_l2: -0.6161 Episode_Reward/joint_torques_l2: -0.1701 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0365 Episode_Reward/unnecessary_wheel_lift: -0.1557 Episode_Reward/rolling_leg_motion: -0.8582 Episode_Reward/obstacle_progress: +2.3090 Episode_Reward/wheel_body_intrusion: -0.3556 Episode_Reward/pair_body_intrusion: -0.2562 Curriculum/terrain_level_mean: +1.5979 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1022 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5918 Episode_Termination/fell_over: 0.0612 Episode_Termination/course_exit: 0.3418 Episode_Termination/illegal_nonwheel_contact: 0.0153 ============================================================================== ============================================================================== Learning iteration 7450/50000 Total steps: 30,515,200 Steps per second: 8,679 Iteration time: 0.472s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0855 Mean episode length: 1443.8 Actor loss: -9.3989 Critic loss: +0.0073 World model loss: +1.2130 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3944 Episode_Reward/track_angular_velocity: +54.5123 Episode_Reward/upright: +28.1089 Episode_Reward/leg_posture: -3.1752 Episode_Reward/standstill: -6.3288 Episode_Reward/action_rate_l2: -0.6225 Episode_Reward/joint_torques_l2: -0.1718 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0310 Episode_Reward/unnecessary_wheel_lift: -0.1422 Episode_Reward/rolling_leg_motion: -0.8614 Episode_Reward/obstacle_progress: +2.3349 Episode_Reward/wheel_body_intrusion: -0.3556 Episode_Reward/pair_body_intrusion: -0.2699 Curriculum/terrain_level_mean: +1.6077 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1009 Metrics/base_velocity/error_vel_yaw: 0.1375 Episode_Termination/time_out: 0.6219 Episode_Termination/fell_over: 0.0348 Episode_Termination/course_exit: 0.3284 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 7500/50000 Total steps: 30,720,000 Steps per second: 8,705 Iteration time: 0.471s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0853 Mean episode length: 1464.2 Actor loss: -9.4413 Critic loss: +0.0024 World model loss: +1.2290 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0311 Episode_Reward/track_angular_velocity: +55.3384 Episode_Reward/upright: +28.4828 Episode_Reward/leg_posture: -3.1489 Episode_Reward/standstill: -6.6255 Episode_Reward/action_rate_l2: -0.6341 Episode_Reward/joint_torques_l2: -0.1729 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0229 Episode_Reward/unnecessary_wheel_lift: -0.1576 Episode_Reward/rolling_leg_motion: -0.8716 Episode_Reward/obstacle_progress: +2.3808 Episode_Reward/wheel_body_intrusion: -0.3492 Episode_Reward/pair_body_intrusion: -0.2871 Curriculum/terrain_level_mean: +1.6179 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1012 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.6505 Episode_Termination/fell_over: 0.0323 Episode_Termination/course_exit: 0.3065 Episode_Termination/illegal_nonwheel_contact: 0.0215 ============================================================================== ============================================================================== Learning iteration 7550/50000 Total steps: 30,924,800 Steps per second: 8,730 Iteration time: 0.469s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0852 Mean episode length: 1402.3 Actor loss: -9.4431 Critic loss: +0.0030 World model loss: +1.1840 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7715 Episode_Reward/track_angular_velocity: +52.9433 Episode_Reward/upright: +27.2467 Episode_Reward/leg_posture: -3.0322 Episode_Reward/standstill: -6.2501 Episode_Reward/action_rate_l2: -0.6094 Episode_Reward/joint_torques_l2: -0.1664 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0356 Episode_Reward/unnecessary_wheel_lift: -0.1416 Episode_Reward/rolling_leg_motion: -0.8392 Episode_Reward/obstacle_progress: +2.2598 Episode_Reward/wheel_body_intrusion: -0.3544 Episode_Reward/pair_body_intrusion: -0.2604 Curriculum/terrain_level_mean: +1.6267 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.5812 Episode_Termination/fell_over: 0.0576 Episode_Termination/course_exit: 0.3508 Episode_Termination/illegal_nonwheel_contact: 0.0157 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2229] ============================================================================== Learning iteration 7600/50000 Total steps: 31,129,600 Steps per second: 8,756 Iteration time: 0.468s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0853 Mean episode length: 1397.0 Actor loss: -9.4354 Critic loss: +0.0034 World model loss: +1.1513 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6001 Episode_Reward/track_angular_velocity: +52.7536 Episode_Reward/upright: +27.1303 Episode_Reward/leg_posture: -3.0680 Episode_Reward/standstill: -6.1049 Episode_Reward/action_rate_l2: -0.6018 Episode_Reward/joint_torques_l2: -0.1674 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.1316 Episode_Reward/rolling_leg_motion: -0.8334 Episode_Reward/obstacle_progress: +2.2732 Episode_Reward/wheel_body_intrusion: -0.3595 Episode_Reward/pair_body_intrusion: -0.2936 Curriculum/terrain_level_mean: +1.6365 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1024 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.5617 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.3889 Episode_Termination/illegal_nonwheel_contact: 0.0185 ============================================================================== ============================================================================== Learning iteration 7650/50000 Total steps: 31,334,400 Steps per second: 8,781 Iteration time: 0.466s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0850 Mean episode length: 1372.8 Actor loss: -9.4370 Critic loss: +0.0040 World model loss: +1.2038 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7058 Episode_Reward/track_angular_velocity: +51.7922 Episode_Reward/upright: +26.6492 Episode_Reward/leg_posture: -2.9986 Episode_Reward/standstill: -6.3183 Episode_Reward/action_rate_l2: -0.5937 Episode_Reward/joint_torques_l2: -0.1654 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.1350 Episode_Reward/rolling_leg_motion: -0.8049 Episode_Reward/obstacle_progress: +2.2464 Episode_Reward/wheel_body_intrusion: -0.3486 Episode_Reward/pair_body_intrusion: -0.3072 Curriculum/terrain_level_mean: +1.6448 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1384 Episode_Termination/time_out: 0.5223 Episode_Termination/fell_over: 0.0446 Episode_Termination/course_exit: 0.4204 Episode_Termination/illegal_nonwheel_contact: 0.0510 ============================================================================== ============================================================================== Learning iteration 7700/50000 Total steps: 31,539,200 Steps per second: 8,806 Iteration time: 0.465s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0848 Mean episode length: 1359.6 Actor loss: -9.4124 Critic loss: +0.0034 World model loss: +1.1549 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1988 Episode_Reward/track_angular_velocity: +51.2570 Episode_Reward/upright: +26.3816 Episode_Reward/leg_posture: -2.9665 Episode_Reward/standstill: -6.3957 Episode_Reward/action_rate_l2: -0.5744 Episode_Reward/joint_torques_l2: -0.1641 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0425 Episode_Reward/unnecessary_wheel_lift: -0.1319 Episode_Reward/rolling_leg_motion: -0.8263 Episode_Reward/obstacle_progress: +2.1969 Episode_Reward/wheel_body_intrusion: -0.3542 Episode_Reward/pair_body_intrusion: -0.3025 Curriculum/terrain_level_mean: +1.6519 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1036 Metrics/base_velocity/error_vel_yaw: 0.1393 Episode_Termination/time_out: 0.5600 Episode_Termination/fell_over: 0.0800 Episode_Termination/course_exit: 0.3486 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1051] ============================================================================== Learning iteration 7750/50000 Total steps: 31,744,000 Steps per second: 8,831 Iteration time: 0.464s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0849 Mean episode length: 1315.1 Actor loss: -9.4148 Critic loss: +0.0045 World model loss: +1.1778 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6537 Episode_Reward/track_angular_velocity: +49.6006 Episode_Reward/upright: +25.5351 Episode_Reward/leg_posture: -2.8867 Episode_Reward/standstill: -6.0314 Episode_Reward/action_rate_l2: -0.5539 Episode_Reward/joint_torques_l2: -0.1582 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0283 Episode_Reward/unnecessary_wheel_lift: -0.1314 Episode_Reward/rolling_leg_motion: -0.7866 Episode_Reward/obstacle_progress: +2.0906 Episode_Reward/wheel_body_intrusion: -0.3270 Episode_Reward/pair_body_intrusion: -0.2804 Curriculum/terrain_level_mean: +1.6609 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1032 Metrics/base_velocity/error_vel_yaw: 0.1389 Episode_Termination/time_out: 0.4882 Episode_Termination/fell_over: 0.0529 Episode_Termination/course_exit: 0.4471 Episode_Termination/illegal_nonwheel_contact: 0.0235 ============================================================================== ============================================================================== Learning iteration 7800/50000 Total steps: 31,948,800 Steps per second: 8,856 Iteration time: 0.463s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0853 Mean episode length: 1274.9 Actor loss: -9.4131 Critic loss: +0.0102 World model loss: +1.1577 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.3789 Episode_Reward/track_angular_velocity: +48.1885 Episode_Reward/upright: +24.7881 Episode_Reward/leg_posture: -2.7696 Episode_Reward/standstill: -5.7223 Episode_Reward/action_rate_l2: -0.5378 Episode_Reward/joint_torques_l2: -0.1514 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0324 Episode_Reward/unnecessary_wheel_lift: -0.1166 Episode_Reward/rolling_leg_motion: -0.7642 Episode_Reward/obstacle_progress: +2.0202 Episode_Reward/wheel_body_intrusion: -0.3111 Episode_Reward/pair_body_intrusion: -0.2543 Curriculum/terrain_level_mean: +1.6694 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1020 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4067 Episode_Termination/fell_over: 0.0733 Episode_Termination/course_exit: 0.5067 Episode_Termination/illegal_nonwheel_contact: 0.0533 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1288] ============================================================================== Learning iteration 7850/50000 Total steps: 32,153,600 Steps per second: 8,881 Iteration time: 0.461s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0853 Mean episode length: 1275.8 Actor loss: -9.3975 Critic loss: +0.0040 World model loss: +1.1332 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.3067 Episode_Reward/track_angular_velocity: +48.2504 Episode_Reward/upright: +24.8133 Episode_Reward/leg_posture: -2.7291 Episode_Reward/standstill: -5.7623 Episode_Reward/action_rate_l2: -0.5535 Episode_Reward/joint_torques_l2: -0.1503 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0299 Episode_Reward/unnecessary_wheel_lift: -0.1105 Episode_Reward/rolling_leg_motion: -0.7600 Episode_Reward/obstacle_progress: +2.0281 Episode_Reward/wheel_body_intrusion: -0.3179 Episode_Reward/pair_body_intrusion: -0.2182 Curriculum/terrain_level_mean: +1.6750 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4500 Episode_Termination/fell_over: 0.0643 Episode_Termination/course_exit: 0.4643 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== ============================================================================== Learning iteration 7900/50000 Total steps: 32,358,400 Steps per second: 8,905 Iteration time: 0.460s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0854 Mean episode length: 1245.7 Actor loss: -9.4068 Critic loss: +0.0025 World model loss: +1.1452 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.2933 Episode_Reward/track_angular_velocity: +47.1250 Episode_Reward/upright: +24.2301 Episode_Reward/leg_posture: -2.6515 Episode_Reward/standstill: -5.6674 Episode_Reward/action_rate_l2: -0.5362 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0170 Episode_Reward/unnecessary_wheel_lift: -0.1032 Episode_Reward/rolling_leg_motion: -0.7335 Episode_Reward/obstacle_progress: +2.0512 Episode_Reward/wheel_body_intrusion: -0.2831 Episode_Reward/pair_body_intrusion: -0.2191 Curriculum/terrain_level_mean: +1.6829 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4478 Episode_Termination/fell_over: 0.0448 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0149 ============================================================================== ============================================================================== Learning iteration 7950/50000 Total steps: 32,563,200 Steps per second: 8,930 Iteration time: 0.459s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0852 Mean episode length: 1258.9 Actor loss: -9.4027 Critic loss: +0.0034 World model loss: +1.1766 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.6899 Episode_Reward/track_angular_velocity: +47.5865 Episode_Reward/upright: +24.4946 Episode_Reward/leg_posture: -2.6731 Episode_Reward/standstill: -5.8645 Episode_Reward/action_rate_l2: -0.5366 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1178 Episode_Reward/rolling_leg_motion: -0.7443 Episode_Reward/obstacle_progress: +2.0743 Episode_Reward/wheel_body_intrusion: -0.3079 Episode_Reward/pair_body_intrusion: -0.2298 Curriculum/terrain_level_mean: +1.6938 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4257 Episode_Termination/fell_over: 0.0541 Episode_Termination/course_exit: 0.5135 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== ============================================================================== Learning iteration 8000/50000 Total steps: 32,768,000 Steps per second: 8,954 Iteration time: 0.457s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0851 Mean episode length: 1351.5 Actor loss: -9.3824 Critic loss: +0.0033 World model loss: +1.1847 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0393 Episode_Reward/track_angular_velocity: +51.0541 Episode_Reward/upright: +26.2937 Episode_Reward/leg_posture: -2.8965 Episode_Reward/standstill: -6.3782 Episode_Reward/action_rate_l2: -0.5692 Episode_Reward/joint_torques_l2: -0.1605 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1213 Episode_Reward/rolling_leg_motion: -0.7803 Episode_Reward/obstacle_progress: +2.1923 Episode_Reward/wheel_body_intrusion: -0.3335 Episode_Reward/pair_body_intrusion: -0.2481 Curriculum/terrain_level_mean: +1.7014 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1021 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4828 Episode_Termination/fell_over: 0.0276 Episode_Termination/course_exit: 0.4759 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 8050/50000 Total steps: 32,972,800 Steps per second: 8,978 Iteration time: 0.456s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0852 Mean episode length: 1322.0 Actor loss: -9.4284 Critic loss: +0.0029 World model loss: +1.1413 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8950 Episode_Reward/track_angular_velocity: +49.9375 Episode_Reward/upright: +25.7277 Episode_Reward/leg_posture: -2.8078 Episode_Reward/standstill: -5.9718 Episode_Reward/action_rate_l2: -0.5678 Episode_Reward/joint_torques_l2: -0.1568 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.1226 Episode_Reward/rolling_leg_motion: -0.7575 Episode_Reward/obstacle_progress: +2.1576 Episode_Reward/wheel_body_intrusion: -0.3221 Episode_Reward/pair_body_intrusion: -0.2951 Curriculum/terrain_level_mean: +1.7092 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4776 Episode_Termination/fell_over: 0.0597 Episode_Termination/course_exit: 0.4627 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 8100/50000 Total steps: 33,177,600 Steps per second: 9,001 Iteration time: 0.455s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1323.9 Actor loss: -9.4184 Critic loss: +0.0028 World model loss: +1.1443 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9810 Episode_Reward/track_angular_velocity: +50.0343 Episode_Reward/upright: +25.7733 Episode_Reward/leg_posture: -2.8536 Episode_Reward/standstill: -6.2065 Episode_Reward/action_rate_l2: -0.5730 Episode_Reward/joint_torques_l2: -0.1573 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.1177 Episode_Reward/rolling_leg_motion: -0.7589 Episode_Reward/obstacle_progress: +2.1407 Episode_Reward/wheel_body_intrusion: -0.3741 Episode_Reward/pair_body_intrusion: -0.3161 Curriculum/terrain_level_mean: +1.7183 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4526 Episode_Termination/fell_over: 0.0438 Episode_Termination/course_exit: 0.4891 Episode_Termination/illegal_nonwheel_contact: 0.0438 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_008100.mp4 ============================================================================== Learning iteration 8150/50000 Total steps: 33,382,400 Steps per second: 9,011 Iteration time: 0.455s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0851 Mean episode length: 1294.1 Actor loss: -9.3942 Critic loss: +0.0067 World model loss: +1.1927 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8985 Episode_Reward/track_angular_velocity: +48.9311 Episode_Reward/upright: +25.1847 Episode_Reward/leg_posture: -2.8376 Episode_Reward/standstill: -5.7784 Episode_Reward/action_rate_l2: -0.5529 Episode_Reward/joint_torques_l2: -0.1541 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0253 Episode_Reward/unnecessary_wheel_lift: -0.1329 Episode_Reward/rolling_leg_motion: -0.7393 Episode_Reward/obstacle_progress: +2.0632 Episode_Reward/wheel_body_intrusion: -0.3693 Episode_Reward/pair_body_intrusion: -0.2951 Curriculum/terrain_level_mean: +1.7251 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4351 Episode_Termination/fell_over: 0.0611 Episode_Termination/course_exit: 0.4885 Episode_Termination/illegal_nonwheel_contact: 0.0611 ============================================================================== ============================================================================== Learning iteration 8200/50000 Total steps: 33,587,200 Steps per second: 9,034 Iteration time: 0.453s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0853 Mean episode length: 1317.9 Actor loss: -9.4301 Critic loss: +0.0035 World model loss: +1.1375 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7445 Episode_Reward/track_angular_velocity: +49.9383 Episode_Reward/upright: +25.6501 Episode_Reward/leg_posture: -2.9043 Episode_Reward/standstill: -5.8683 Episode_Reward/action_rate_l2: -0.5584 Episode_Reward/joint_torques_l2: -0.1558 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0309 Episode_Reward/unnecessary_wheel_lift: -0.1212 Episode_Reward/rolling_leg_motion: -0.7491 Episode_Reward/obstacle_progress: +2.1234 Episode_Reward/wheel_body_intrusion: -0.3689 Episode_Reward/pair_body_intrusion: -0.3151 Curriculum/terrain_level_mean: +1.7324 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4776 Episode_Termination/fell_over: 0.0672 Episode_Termination/course_exit: 0.4328 Episode_Termination/illegal_nonwheel_contact: 0.0597 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1109] ============================================================================== Learning iteration 8250/50000 Total steps: 33,792,000 Steps per second: 9,057 Iteration time: 0.452s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0852 Mean episode length: 1282.1 Actor loss: -9.4145 Critic loss: +0.0028 World model loss: +1.1247 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.3546 Episode_Reward/track_angular_velocity: +48.4950 Episode_Reward/upright: +24.9582 Episode_Reward/leg_posture: -2.7776 Episode_Reward/standstill: -5.6310 Episode_Reward/action_rate_l2: -0.5583 Episode_Reward/joint_torques_l2: -0.1513 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0250 Episode_Reward/unnecessary_wheel_lift: -0.1330 Episode_Reward/rolling_leg_motion: -0.7388 Episode_Reward/obstacle_progress: +2.0618 Episode_Reward/wheel_body_intrusion: -0.3421 Episode_Reward/pair_body_intrusion: -0.2772 Curriculum/terrain_level_mean: +1.7400 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1044 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4274 Episode_Termination/fell_over: 0.0403 Episode_Termination/course_exit: 0.4919 Episode_Termination/illegal_nonwheel_contact: 0.0645 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[842] RECOVERY reset 1 invalid environments; first IDs=[466] ============================================================================== Learning iteration 8300/50000 Total steps: 33,996,800 Steps per second: 9,080 Iteration time: 0.451s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0852 Mean episode length: 1333.9 Actor loss: -9.3846 Critic loss: +0.0031 World model loss: +1.1215 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1996 Episode_Reward/track_angular_velocity: +50.4783 Episode_Reward/upright: +25.9865 Episode_Reward/leg_posture: -2.9120 Episode_Reward/standstill: -5.8803 Episode_Reward/action_rate_l2: -0.5711 Episode_Reward/joint_torques_l2: -0.1582 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0287 Episode_Reward/unnecessary_wheel_lift: -0.1218 Episode_Reward/rolling_leg_motion: -0.7852 Episode_Reward/obstacle_progress: +2.0672 Episode_Reward/wheel_body_intrusion: -0.3756 Episode_Reward/pair_body_intrusion: -0.2921 Curriculum/terrain_level_mean: +1.7468 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4832 Episode_Termination/fell_over: 0.0537 Episode_Termination/course_exit: 0.4430 Episode_Termination/illegal_nonwheel_contact: 0.0268 ============================================================================== ============================================================================== Learning iteration 8350/50000 Total steps: 34,201,600 Steps per second: 9,102 Iteration time: 0.450s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0851 Mean episode length: 1332.5 Actor loss: -9.4059 Critic loss: +0.0226 World model loss: +1.1053 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0670 Episode_Reward/track_angular_velocity: +50.4676 Episode_Reward/upright: +25.9431 Episode_Reward/leg_posture: -2.9079 Episode_Reward/standstill: -5.9554 Episode_Reward/action_rate_l2: -0.5724 Episode_Reward/joint_torques_l2: -0.1582 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0286 Episode_Reward/unnecessary_wheel_lift: -0.1249 Episode_Reward/rolling_leg_motion: -0.7818 Episode_Reward/obstacle_progress: +2.1008 Episode_Reward/wheel_body_intrusion: -0.3472 Episode_Reward/pair_body_intrusion: -0.3134 Curriculum/terrain_level_mean: +1.7534 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4514 Episode_Termination/fell_over: 0.0556 Episode_Termination/course_exit: 0.4722 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1570] ============================================================================== Learning iteration 8400/50000 Total steps: 34,406,400 Steps per second: 9,125 Iteration time: 0.449s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0847 Mean episode length: 1251.0 Actor loss: -9.3898 Critic loss: +0.0054 World model loss: +1.1679 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.2233 Episode_Reward/track_angular_velocity: +47.3479 Episode_Reward/upright: +24.3314 Episode_Reward/leg_posture: -2.7518 Episode_Reward/standstill: -5.9682 Episode_Reward/action_rate_l2: -0.5436 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0147 Episode_Reward/unnecessary_wheel_lift: -0.1320 Episode_Reward/rolling_leg_motion: -0.7411 Episode_Reward/obstacle_progress: +1.9436 Episode_Reward/wheel_body_intrusion: -0.3133 Episode_Reward/pair_body_intrusion: -0.2658 Curriculum/terrain_level_mean: +1.7661 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.3688 Episode_Termination/fell_over: 0.0250 Episode_Termination/course_exit: 0.5938 Episode_Termination/illegal_nonwheel_contact: 0.0187 ============================================================================== ============================================================================== Learning iteration 8450/50000 Total steps: 34,611,200 Steps per second: 9,148 Iteration time: 0.448s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0852 Mean episode length: 1311.1 Actor loss: -9.4275 Critic loss: +0.0026 World model loss: +1.1436 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.3749 Episode_Reward/track_angular_velocity: +49.6266 Episode_Reward/upright: +25.5140 Episode_Reward/leg_posture: -2.8998 Episode_Reward/standstill: -5.7510 Episode_Reward/action_rate_l2: -0.5626 Episode_Reward/joint_torques_l2: -0.1575 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0179 Episode_Reward/unnecessary_wheel_lift: -0.1394 Episode_Reward/rolling_leg_motion: -0.7668 Episode_Reward/obstacle_progress: +2.0423 Episode_Reward/wheel_body_intrusion: -0.3309 Episode_Reward/pair_body_intrusion: -0.2824 Curriculum/terrain_level_mean: +1.7729 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4211 Episode_Termination/fell_over: 0.0526 Episode_Termination/course_exit: 0.5263 Episode_Termination/illegal_nonwheel_contact: 0.0226 ============================================================================== ============================================================================== Learning iteration 8500/50000 Total steps: 34,816,000 Steps per second: 9,170 Iteration time: 0.447s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0847 Mean episode length: 1209.8 Actor loss: -9.4120 Critic loss: +0.0043 World model loss: +1.0839 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +41.8205 Episode_Reward/track_angular_velocity: +45.7830 Episode_Reward/upright: +23.5339 Episode_Reward/leg_posture: -2.6446 Episode_Reward/standstill: -5.8565 Episode_Reward/action_rate_l2: -0.5294 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0331 Episode_Reward/unnecessary_wheel_lift: -0.1425 Episode_Reward/rolling_leg_motion: -0.7007 Episode_Reward/obstacle_progress: +1.9055 Episode_Reward/wheel_body_intrusion: -0.3045 Episode_Reward/pair_body_intrusion: -0.2503 Curriculum/terrain_level_mean: +1.7820 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.3582 Episode_Termination/fell_over: 0.0896 Episode_Termination/course_exit: 0.5373 Episode_Termination/illegal_nonwheel_contact: 0.0672 ============================================================================== ============================================================================== Learning iteration 8550/50000 Total steps: 35,020,800 Steps per second: 9,193 Iteration time: 0.446s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1313.4 Actor loss: -9.3726 Critic loss: +0.0048 World model loss: +1.0885 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5039 Episode_Reward/track_angular_velocity: +49.6615 Episode_Reward/upright: +25.5192 Episode_Reward/leg_posture: -2.8055 Episode_Reward/standstill: -6.2243 Episode_Reward/action_rate_l2: -0.5807 Episode_Reward/joint_torques_l2: -0.1589 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0333 Episode_Reward/unnecessary_wheel_lift: -0.1275 Episode_Reward/rolling_leg_motion: -0.7605 Episode_Reward/obstacle_progress: +2.2001 Episode_Reward/wheel_body_intrusion: -0.3495 Episode_Reward/pair_body_intrusion: -0.3314 Curriculum/terrain_level_mean: +1.7893 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1036 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4331 Episode_Termination/fell_over: 0.0637 Episode_Termination/course_exit: 0.4841 Episode_Termination/illegal_nonwheel_contact: 0.0446 ============================================================================== ============================================================================== Learning iteration 8600/50000 Total steps: 35,225,600 Steps per second: 9,215 Iteration time: 0.444s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0845 Mean episode length: 1322.8 Actor loss: -9.3953 Critic loss: +0.0059 World model loss: +1.1142 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6484 Episode_Reward/track_angular_velocity: +49.9362 Episode_Reward/upright: +25.6883 Episode_Reward/leg_posture: -2.7762 Episode_Reward/standstill: -6.7027 Episode_Reward/action_rate_l2: -0.5847 Episode_Reward/joint_torques_l2: -0.1607 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.1281 Episode_Reward/rolling_leg_motion: -0.7582 Episode_Reward/obstacle_progress: +2.2243 Episode_Reward/wheel_body_intrusion: -0.3442 Episode_Reward/pair_body_intrusion: -0.2983 Curriculum/terrain_level_mean: +1.7954 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1053 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.4685 Episode_Termination/fell_over: 0.0699 Episode_Termination/course_exit: 0.4545 Episode_Termination/illegal_nonwheel_contact: 0.0280 ============================================================================== ============================================================================== Learning iteration 8650/50000 Total steps: 35,430,400 Steps per second: 9,237 Iteration time: 0.443s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1266.6 Actor loss: -9.4039 Critic loss: +0.0064 World model loss: +1.0633 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.8513 Episode_Reward/track_angular_velocity: +47.9165 Episode_Reward/upright: +24.6050 Episode_Reward/leg_posture: -2.6723 Episode_Reward/standstill: -6.1476 Episode_Reward/action_rate_l2: -0.5490 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0223 Episode_Reward/unnecessary_wheel_lift: -0.1249 Episode_Reward/rolling_leg_motion: -0.7198 Episode_Reward/obstacle_progress: +2.1203 Episode_Reward/wheel_body_intrusion: -0.3139 Episode_Reward/pair_body_intrusion: -0.2651 Curriculum/terrain_level_mean: +1.8066 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1041 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4610 Episode_Termination/fell_over: 0.0390 Episode_Termination/course_exit: 0.4805 Episode_Termination/illegal_nonwheel_contact: 0.0325 ============================================================================== ============================================================================== Learning iteration 8700/50000 Total steps: 35,635,200 Steps per second: 9,259 Iteration time: 0.442s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1318.5 Actor loss: -9.4100 Critic loss: +0.0047 World model loss: +1.0930 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5608 Episode_Reward/track_angular_velocity: +49.9002 Episode_Reward/upright: +25.6111 Episode_Reward/leg_posture: -2.7881 Episode_Reward/standstill: -6.5796 Episode_Reward/action_rate_l2: -0.5713 Episode_Reward/joint_torques_l2: -0.1613 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0180 Episode_Reward/unnecessary_wheel_lift: -0.1371 Episode_Reward/rolling_leg_motion: -0.7535 Episode_Reward/obstacle_progress: +2.1996 Episode_Reward/wheel_body_intrusion: -0.3124 Episode_Reward/pair_body_intrusion: -0.2666 Curriculum/terrain_level_mean: +1.8171 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4395 Episode_Termination/fell_over: 0.0446 Episode_Termination/course_exit: 0.5159 Episode_Termination/illegal_nonwheel_contact: 0.0064 ============================================================================== ============================================================================== Learning iteration 8750/50000 Total steps: 35,840,000 Steps per second: 9,281 Iteration time: 0.441s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0842 Mean episode length: 1298.3 Actor loss: -9.4196 Critic loss: +0.0030 World model loss: +1.0823 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.9471 Episode_Reward/track_angular_velocity: +49.0726 Episode_Reward/upright: +25.2125 Episode_Reward/leg_posture: -2.7385 Episode_Reward/standstill: -7.0230 Episode_Reward/action_rate_l2: -0.5621 Episode_Reward/joint_torques_l2: -0.1578 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0406 Episode_Reward/unnecessary_wheel_lift: -0.1204 Episode_Reward/rolling_leg_motion: -0.7468 Episode_Reward/obstacle_progress: +2.1537 Episode_Reward/wheel_body_intrusion: -0.3294 Episode_Reward/pair_body_intrusion: -0.2953 Curriculum/terrain_level_mean: +1.8210 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4379 Episode_Termination/fell_over: 0.0784 Episode_Termination/course_exit: 0.4575 Episode_Termination/illegal_nonwheel_contact: 0.0654 ============================================================================== ============================================================================== Learning iteration 8800/50000 Total steps: 36,044,800 Steps per second: 9,303 Iteration time: 0.440s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0846 Mean episode length: 1329.0 Actor loss: -9.3783 Critic loss: +0.0049 World model loss: +1.0921 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9315 Episode_Reward/track_angular_velocity: +50.2863 Episode_Reward/upright: +25.8042 Episode_Reward/leg_posture: -2.7966 Episode_Reward/standstill: -6.6913 Episode_Reward/action_rate_l2: -0.5790 Episode_Reward/joint_torques_l2: -0.1613 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0337 Episode_Reward/unnecessary_wheel_lift: -0.1257 Episode_Reward/rolling_leg_motion: -0.7537 Episode_Reward/obstacle_progress: +2.1404 Episode_Reward/wheel_body_intrusion: -0.3211 Episode_Reward/pair_body_intrusion: -0.2802 Curriculum/terrain_level_mean: +1.8281 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1050 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4531 Episode_Termination/fell_over: 0.0859 Episode_Termination/course_exit: 0.4453 Episode_Termination/illegal_nonwheel_contact: 0.0391 ============================================================================== ============================================================================== Learning iteration 8850/50000 Total steps: 36,249,600 Steps per second: 9,324 Iteration time: 0.439s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0841 Mean episode length: 1317.2 Actor loss: -9.4123 Critic loss: +0.0029 World model loss: +1.0741 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5337 Episode_Reward/track_angular_velocity: +49.8068 Episode_Reward/upright: +25.5795 Episode_Reward/leg_posture: -2.7826 Episode_Reward/standstill: -7.2234 Episode_Reward/action_rate_l2: -0.5610 Episode_Reward/joint_torques_l2: -0.1590 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0334 Episode_Reward/unnecessary_wheel_lift: -0.1492 Episode_Reward/rolling_leg_motion: -0.7527 Episode_Reward/obstacle_progress: +2.1664 Episode_Reward/wheel_body_intrusion: -0.3372 Episode_Reward/pair_body_intrusion: -0.2929 Curriculum/terrain_level_mean: +1.8340 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4354 Episode_Termination/fell_over: 0.0612 Episode_Termination/course_exit: 0.4762 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 8900/50000 Total steps: 36,454,400 Steps per second: 9,345 Iteration time: 0.438s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1385.4 Actor loss: -9.3705 Critic loss: +0.0167 World model loss: +1.1001 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9479 Episode_Reward/track_angular_velocity: +52.3954 Episode_Reward/upright: +26.9401 Episode_Reward/leg_posture: -2.9058 Episode_Reward/standstill: -7.1553 Episode_Reward/action_rate_l2: -0.5899 Episode_Reward/joint_torques_l2: -0.1669 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0325 Episode_Reward/unnecessary_wheel_lift: -0.1379 Episode_Reward/rolling_leg_motion: -0.7886 Episode_Reward/obstacle_progress: +2.3018 Episode_Reward/wheel_body_intrusion: -0.3693 Episode_Reward/pair_body_intrusion: -0.2837 Curriculum/terrain_level_mean: +1.8416 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5185 Episode_Termination/fell_over: 0.0617 Episode_Termination/course_exit: 0.4074 Episode_Termination/illegal_nonwheel_contact: 0.0247 ============================================================================== ============================================================================== Learning iteration 8950/50000 Total steps: 36,659,200 Steps per second: 9,367 Iteration time: 0.437s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1339.2 Actor loss: -9.3520 Critic loss: +0.0089 World model loss: +1.0863 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2383 Episode_Reward/track_angular_velocity: +50.6495 Episode_Reward/upright: +26.0369 Episode_Reward/leg_posture: -2.8589 Episode_Reward/standstill: -6.4734 Episode_Reward/action_rate_l2: -0.5759 Episode_Reward/joint_torques_l2: -0.1609 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0314 Episode_Reward/unnecessary_wheel_lift: -0.1453 Episode_Reward/rolling_leg_motion: -0.7518 Episode_Reward/obstacle_progress: +2.2341 Episode_Reward/wheel_body_intrusion: -0.3579 Episode_Reward/pair_body_intrusion: -0.2997 Curriculum/terrain_level_mean: +1.8484 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1050 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5035 Episode_Termination/fell_over: 0.0638 Episode_Termination/course_exit: 0.4113 Episode_Termination/illegal_nonwheel_contact: 0.0567 ============================================================================== ============================================================================== Learning iteration 9000/50000 Total steps: 36,864,000 Steps per second: 9,387 Iteration time: 0.436s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0846 Mean episode length: 1367.9 Actor loss: -9.3708 Critic loss: +0.0041 World model loss: +1.0923 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2934 Episode_Reward/track_angular_velocity: +51.7639 Episode_Reward/upright: +26.6081 Episode_Reward/leg_posture: -2.8997 Episode_Reward/standstill: -6.9565 Episode_Reward/action_rate_l2: -0.5985 Episode_Reward/joint_torques_l2: -0.1626 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0321 Episode_Reward/unnecessary_wheel_lift: -0.1371 Episode_Reward/rolling_leg_motion: -0.7738 Episode_Reward/obstacle_progress: +2.2786 Episode_Reward/wheel_body_intrusion: -0.3321 Episode_Reward/pair_body_intrusion: -0.3271 Curriculum/terrain_level_mean: +1.8550 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1046 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5287 Episode_Termination/fell_over: 0.0573 Episode_Termination/course_exit: 0.3949 Episode_Termination/illegal_nonwheel_contact: 0.0573 ============================================================================== ============================================================================== Learning iteration 9050/50000 Total steps: 37,068,800 Steps per second: 9,407 Iteration time: 0.435s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0840 Mean episode length: 1367.7 Actor loss: -9.4020 Critic loss: +0.0044 World model loss: +1.0637 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2590 Episode_Reward/track_angular_velocity: +51.7825 Episode_Reward/upright: +26.5982 Episode_Reward/leg_posture: -2.9394 Episode_Reward/standstill: -7.4296 Episode_Reward/action_rate_l2: -0.5994 Episode_Reward/joint_torques_l2: -0.1619 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0427 Episode_Reward/unnecessary_wheel_lift: -0.1641 Episode_Reward/rolling_leg_motion: -0.8034 Episode_Reward/obstacle_progress: +2.1850 Episode_Reward/wheel_body_intrusion: -0.3576 Episode_Reward/pair_body_intrusion: -0.3912 Curriculum/terrain_level_mean: +1.8616 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.5444 Episode_Termination/fell_over: 0.0667 Episode_Termination/course_exit: 0.3667 Episode_Termination/illegal_nonwheel_contact: 0.0389 ============================================================================== ============================================================================== Learning iteration 9100/50000 Total steps: 37,273,600 Steps per second: 9,428 Iteration time: 0.434s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0845 Mean episode length: 1377.8 Actor loss: -9.3669 Critic loss: +0.0247 World model loss: +1.1051 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7023 Episode_Reward/track_angular_velocity: +52.1603 Episode_Reward/upright: +26.8236 Episode_Reward/leg_posture: -2.9659 Episode_Reward/standstill: -7.0721 Episode_Reward/action_rate_l2: -0.5970 Episode_Reward/joint_torques_l2: -0.1635 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0323 Episode_Reward/unnecessary_wheel_lift: -0.1469 Episode_Reward/rolling_leg_motion: -0.8043 Episode_Reward/obstacle_progress: +2.2200 Episode_Reward/wheel_body_intrusion: -0.3560 Episode_Reward/pair_body_intrusion: -0.3631 Curriculum/terrain_level_mean: +1.8682 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1035 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5385 Episode_Termination/fell_over: 0.0296 Episode_Termination/course_exit: 0.3905 Episode_Termination/illegal_nonwheel_contact: 0.0651 ============================================================================== ============================================================================== Learning iteration 9150/50000 Total steps: 37,478,400 Steps per second: 9,449 Iteration time: 0.433s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0845 Mean episode length: 1373.6 Actor loss: -9.4040 Critic loss: +0.0047 World model loss: +1.0877 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5483 Episode_Reward/track_angular_velocity: +51.9684 Episode_Reward/upright: +26.7363 Episode_Reward/leg_posture: -2.9240 Episode_Reward/standstill: -7.0270 Episode_Reward/action_rate_l2: -0.6000 Episode_Reward/joint_torques_l2: -0.1611 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0349 Episode_Reward/unnecessary_wheel_lift: -0.1450 Episode_Reward/rolling_leg_motion: -0.7990 Episode_Reward/obstacle_progress: +2.2249 Episode_Reward/wheel_body_intrusion: -0.3264 Episode_Reward/pair_body_intrusion: -0.3378 Curriculum/terrain_level_mean: +1.8743 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1038 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.5497 Episode_Termination/fell_over: 0.0585 Episode_Termination/course_exit: 0.3743 Episode_Termination/illegal_nonwheel_contact: 0.0526 ============================================================================== ============================================================================== Learning iteration 9200/50000 Total steps: 37,683,200 Steps per second: 9,469 Iteration time: 0.433s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1486.8 Actor loss: -9.4168 Critic loss: +0.0043 World model loss: +1.1104 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.4177 Episode_Reward/track_angular_velocity: +56.2942 Episode_Reward/upright: +28.9744 Episode_Reward/leg_posture: -3.1568 Episode_Reward/standstill: -7.1015 Episode_Reward/action_rate_l2: -0.6349 Episode_Reward/joint_torques_l2: -0.1738 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0203 Episode_Reward/unnecessary_wheel_lift: -0.1349 Episode_Reward/rolling_leg_motion: -0.8450 Episode_Reward/obstacle_progress: +2.3827 Episode_Reward/wheel_body_intrusion: -0.3392 Episode_Reward/pair_body_intrusion: -0.3316 Curriculum/terrain_level_mean: +1.8796 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1038 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.6264 Episode_Termination/fell_over: 0.0230 Episode_Termination/course_exit: 0.3333 Episode_Termination/illegal_nonwheel_contact: 0.0402 ============================================================================== ============================================================================== Learning iteration 9250/50000 Total steps: 37,888,000 Steps per second: 9,490 Iteration time: 0.432s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1390.4 Actor loss: -9.4040 Critic loss: +0.0044 World model loss: +1.0296 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2036 Episode_Reward/track_angular_velocity: +52.6460 Episode_Reward/upright: +27.0795 Episode_Reward/leg_posture: -2.9437 Episode_Reward/standstill: -6.8837 Episode_Reward/action_rate_l2: -0.6003 Episode_Reward/joint_torques_l2: -0.1640 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0272 Episode_Reward/unnecessary_wheel_lift: -0.1419 Episode_Reward/rolling_leg_motion: -0.7921 Episode_Reward/obstacle_progress: +2.2859 Episode_Reward/wheel_body_intrusion: -0.3360 Episode_Reward/pair_body_intrusion: -0.3319 Curriculum/terrain_level_mean: +1.8877 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5775 Episode_Termination/fell_over: 0.0481 Episode_Termination/course_exit: 0.3690 Episode_Termination/illegal_nonwheel_contact: 0.0321 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1438] ============================================================================== Learning iteration 9300/50000 Total steps: 38,092,800 Steps per second: 9,510 Iteration time: 0.431s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0851 Mean episode length: 1412.9 Actor loss: -9.4108 Critic loss: +0.0039 World model loss: +1.1142 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9204 Episode_Reward/track_angular_velocity: +53.5948 Episode_Reward/upright: +27.5139 Episode_Reward/leg_posture: -3.0428 Episode_Reward/standstill: -6.6329 Episode_Reward/action_rate_l2: -0.5917 Episode_Reward/joint_torques_l2: -0.1658 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0276 Episode_Reward/unnecessary_wheel_lift: -0.1401 Episode_Reward/rolling_leg_motion: -0.7931 Episode_Reward/obstacle_progress: +2.3339 Episode_Reward/wheel_body_intrusion: -0.3416 Episode_Reward/pair_body_intrusion: -0.3592 Curriculum/terrain_level_mean: +1.8936 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1034 Metrics/base_velocity/error_vel_yaw: 0.1314 Episode_Termination/time_out: 0.5636 Episode_Termination/fell_over: 0.0364 Episode_Termination/course_exit: 0.3758 Episode_Termination/illegal_nonwheel_contact: 0.0424 ============================================================================== ============================================================================== Learning iteration 9350/50000 Total steps: 38,297,600 Steps per second: 9,530 Iteration time: 0.430s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0848 Mean episode length: 1352.0 Actor loss: -9.3826 Critic loss: +0.0036 World model loss: +1.0517 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6844 Episode_Reward/track_angular_velocity: +51.2704 Episode_Reward/upright: +26.3304 Episode_Reward/leg_posture: -2.8793 Episode_Reward/standstill: -6.6564 Episode_Reward/action_rate_l2: -0.5781 Episode_Reward/joint_torques_l2: -0.1587 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0502 Episode_Reward/unnecessary_wheel_lift: -0.1590 Episode_Reward/rolling_leg_motion: -0.7724 Episode_Reward/obstacle_progress: +2.1598 Episode_Reward/wheel_body_intrusion: -0.2911 Episode_Reward/pair_body_intrusion: -0.2699 Curriculum/terrain_level_mean: +1.8962 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.5276 Episode_Termination/fell_over: 0.0859 Episode_Termination/course_exit: 0.3558 Episode_Termination/illegal_nonwheel_contact: 0.0613 ============================================================================== ============================================================================== Learning iteration 9400/50000 Total steps: 38,502,400 Steps per second: 9,550 Iteration time: 0.429s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0848 Mean episode length: 1359.2 Actor loss: -9.4220 Critic loss: +0.0026 World model loss: +1.0001 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0615 Episode_Reward/track_angular_velocity: +51.4895 Episode_Reward/upright: +26.4700 Episode_Reward/leg_posture: -2.8909 Episode_Reward/standstill: -6.7863 Episode_Reward/action_rate_l2: -0.5787 Episode_Reward/joint_torques_l2: -0.1591 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0292 Episode_Reward/unnecessary_wheel_lift: -0.1314 Episode_Reward/rolling_leg_motion: -0.7636 Episode_Reward/obstacle_progress: +2.1680 Episode_Reward/wheel_body_intrusion: -0.3154 Episode_Reward/pair_body_intrusion: -0.3147 Curriculum/terrain_level_mean: +1.9033 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1035 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.5122 Episode_Termination/fell_over: 0.0610 Episode_Termination/course_exit: 0.4207 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1276] ============================================================================== Learning iteration 9450/50000 Total steps: 38,707,200 Steps per second: 9,570 Iteration time: 0.428s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1351.3 Actor loss: -9.4160 Critic loss: +0.0079 World model loss: +0.9963 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7933 Episode_Reward/track_angular_velocity: +51.2370 Episode_Reward/upright: +26.3314 Episode_Reward/leg_posture: -2.9015 Episode_Reward/standstill: -6.5647 Episode_Reward/action_rate_l2: -0.5834 Episode_Reward/joint_torques_l2: -0.1583 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0396 Episode_Reward/unnecessary_wheel_lift: -0.1451 Episode_Reward/rolling_leg_motion: -0.7835 Episode_Reward/obstacle_progress: +2.1438 Episode_Reward/wheel_body_intrusion: -0.3303 Episode_Reward/pair_body_intrusion: -0.2982 Curriculum/terrain_level_mean: +1.9106 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1031 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.5256 Episode_Termination/fell_over: 0.0962 Episode_Termination/course_exit: 0.3782 Episode_Termination/illegal_nonwheel_contact: 0.0641 ============================================================================== ============================================================================== Learning iteration 9500/50000 Total steps: 38,912,000 Steps per second: 9,589 Iteration time: 0.427s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0845 Mean episode length: 1397.9 Actor loss: -9.4087 Critic loss: +0.0038 World model loss: +1.0258 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3410 Episode_Reward/track_angular_velocity: +52.9186 Episode_Reward/upright: +27.2229 Episode_Reward/leg_posture: -3.0100 Episode_Reward/standstill: -7.0737 Episode_Reward/action_rate_l2: -0.6096 Episode_Reward/joint_torques_l2: -0.1665 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0300 Episode_Reward/unnecessary_wheel_lift: -0.1259 Episode_Reward/rolling_leg_motion: -0.7879 Episode_Reward/obstacle_progress: +2.1624 Episode_Reward/wheel_body_intrusion: -0.3625 Episode_Reward/pair_body_intrusion: -0.3154 Curriculum/terrain_level_mean: +1.9163 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1039 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5183 Episode_Termination/fell_over: 0.0488 Episode_Termination/course_exit: 0.4146 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== ============================================================================== Learning iteration 9550/50000 Total steps: 39,116,800 Steps per second: 9,608 Iteration time: 0.426s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0849 Mean episode length: 1356.4 Actor loss: -9.3957 Critic loss: +0.0034 World model loss: +0.9875 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9251 Episode_Reward/track_angular_velocity: +51.3697 Episode_Reward/upright: +26.4063 Episode_Reward/leg_posture: -2.9714 Episode_Reward/standstill: -6.4815 Episode_Reward/action_rate_l2: -0.5808 Episode_Reward/joint_torques_l2: -0.1601 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.1102 Episode_Reward/rolling_leg_motion: -0.7737 Episode_Reward/obstacle_progress: +2.1797 Episode_Reward/wheel_body_intrusion: -0.3650 Episode_Reward/pair_body_intrusion: -0.3272 Curriculum/terrain_level_mean: +1.9253 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1032 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5033 Episode_Termination/fell_over: 0.0199 Episode_Termination/course_exit: 0.4570 Episode_Termination/illegal_nonwheel_contact: 0.0265 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1571] ============================================================================== Learning iteration 9600/50000 Total steps: 39,321,600 Steps per second: 9,628 Iteration time: 0.425s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1335.5 Actor loss: -9.4027 Critic loss: +0.0038 World model loss: +1.0201 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3045 Episode_Reward/track_angular_velocity: +50.5824 Episode_Reward/upright: +25.9984 Episode_Reward/leg_posture: -2.9027 Episode_Reward/standstill: -6.7462 Episode_Reward/action_rate_l2: -0.5827 Episode_Reward/joint_torques_l2: -0.1572 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.1335 Episode_Reward/rolling_leg_motion: -0.7539 Episode_Reward/obstacle_progress: +2.1286 Episode_Reward/wheel_body_intrusion: -0.3305 Episode_Reward/pair_body_intrusion: -0.3123 Curriculum/terrain_level_mean: +1.9319 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1025 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4521 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4863 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 9650/50000 Total steps: 39,526,400 Steps per second: 9,647 Iteration time: 0.425s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0850 Mean episode length: 1370.3 Actor loss: -9.3989 Critic loss: +0.0027 World model loss: +1.0063 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4553 Episode_Reward/track_angular_velocity: +51.8870 Episode_Reward/upright: +26.6600 Episode_Reward/leg_posture: -2.9766 Episode_Reward/standstill: -6.4590 Episode_Reward/action_rate_l2: -0.5876 Episode_Reward/joint_torques_l2: -0.1608 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.1178 Episode_Reward/rolling_leg_motion: -0.7539 Episode_Reward/obstacle_progress: +2.1641 Episode_Reward/wheel_body_intrusion: -0.3307 Episode_Reward/pair_body_intrusion: -0.3178 Curriculum/terrain_level_mean: +1.9365 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1030 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4865 Episode_Termination/fell_over: 0.0450 Episode_Termination/course_exit: 0.4505 Episode_Termination/illegal_nonwheel_contact: 0.0541 ============================================================================== ============================================================================== Learning iteration 9700/50000 Total steps: 39,731,200 Steps per second: 9,666 Iteration time: 0.424s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0844 Mean episode length: 1334.7 Actor loss: -9.3926 Critic loss: +0.0085 World model loss: +1.0088 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1514 Episode_Reward/track_angular_velocity: +50.5876 Episode_Reward/upright: +25.9432 Episode_Reward/leg_posture: -2.8777 Episode_Reward/standstill: -7.0067 Episode_Reward/action_rate_l2: -0.5680 Episode_Reward/joint_torques_l2: -0.1559 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0339 Episode_Reward/unnecessary_wheel_lift: -0.1546 Episode_Reward/rolling_leg_motion: -0.7427 Episode_Reward/obstacle_progress: +2.1562 Episode_Reward/wheel_body_intrusion: -0.3201 Episode_Reward/pair_body_intrusion: -0.3636 Curriculum/terrain_level_mean: +1.9421 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1038 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4422 Episode_Termination/fell_over: 0.0748 Episode_Termination/course_exit: 0.4694 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 9750/50000 Total steps: 39,936,000 Steps per second: 9,685 Iteration time: 0.423s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0848 Mean episode length: 1332.6 Actor loss: -9.3913 Critic loss: +0.0044 World model loss: +0.9433 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1245 Episode_Reward/track_angular_velocity: +50.4686 Episode_Reward/upright: +25.9134 Episode_Reward/leg_posture: -2.8196 Episode_Reward/standstill: -6.5779 Episode_Reward/action_rate_l2: -0.5663 Episode_Reward/joint_torques_l2: -0.1562 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0208 Episode_Reward/unnecessary_wheel_lift: -0.1166 Episode_Reward/rolling_leg_motion: -0.7329 Episode_Reward/obstacle_progress: +2.1790 Episode_Reward/wheel_body_intrusion: -0.3451 Episode_Reward/pair_body_intrusion: -0.3261 Curriculum/terrain_level_mean: +1.9521 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5030 Episode_Termination/fell_over: 0.0242 Episode_Termination/course_exit: 0.4485 Episode_Termination/illegal_nonwheel_contact: 0.0364 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2027] ============================================================================== Learning iteration 9800/50000 Total steps: 40,140,800 Steps per second: 9,704 Iteration time: 0.422s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0846 Mean episode length: 1335.6 Actor loss: -9.4047 Critic loss: +0.0024 World model loss: +1.0166 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0616 Episode_Reward/track_angular_velocity: +50.5227 Episode_Reward/upright: +25.9770 Episode_Reward/leg_posture: -2.8629 Episode_Reward/standstill: -6.5720 Episode_Reward/action_rate_l2: -0.5775 Episode_Reward/joint_torques_l2: -0.1576 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1034 Episode_Reward/rolling_leg_motion: -0.7286 Episode_Reward/obstacle_progress: +2.1421 Episode_Reward/wheel_body_intrusion: -0.3313 Episode_Reward/pair_body_intrusion: -0.3571 Curriculum/terrain_level_mean: +1.9570 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4803 Episode_Termination/fell_over: 0.0394 Episode_Termination/course_exit: 0.4646 Episode_Termination/illegal_nonwheel_contact: 0.0157 ============================================================================== ============================================================================== Learning iteration 9850/50000 Total steps: 40,345,600 Steps per second: 9,723 Iteration time: 0.421s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0843 Mean episode length: 1329.2 Actor loss: -9.3961 Critic loss: +0.0028 World model loss: +0.9880 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8552 Episode_Reward/track_angular_velocity: +50.2937 Episode_Reward/upright: +25.8451 Episode_Reward/leg_posture: -2.8290 Episode_Reward/standstill: -6.9189 Episode_Reward/action_rate_l2: -0.5756 Episode_Reward/joint_torques_l2: -0.1559 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0312 Episode_Reward/unnecessary_wheel_lift: -0.1481 Episode_Reward/rolling_leg_motion: -0.7380 Episode_Reward/obstacle_progress: +2.1665 Episode_Reward/wheel_body_intrusion: -0.3190 Episode_Reward/pair_body_intrusion: -0.3580 Curriculum/terrain_level_mean: +1.9636 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1045 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4853 Episode_Termination/fell_over: 0.0588 Episode_Termination/course_exit: 0.4265 Episode_Termination/illegal_nonwheel_contact: 0.0882 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2327] ============================================================================== Learning iteration 9900/50000 Total steps: 40,550,400 Steps per second: 9,742 Iteration time: 0.420s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0843 Mean episode length: 1330.0 Actor loss: -9.3692 Critic loss: +0.0049 World model loss: +1.0221 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8782 Episode_Reward/track_angular_velocity: +50.2806 Episode_Reward/upright: +25.8447 Episode_Reward/leg_posture: -2.8330 Episode_Reward/standstill: -6.8216 Episode_Reward/action_rate_l2: -0.5678 Episode_Reward/joint_torques_l2: -0.1557 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0338 Episode_Reward/unnecessary_wheel_lift: -0.1519 Episode_Reward/rolling_leg_motion: -0.7419 Episode_Reward/obstacle_progress: +2.1459 Episode_Reward/wheel_body_intrusion: -0.3180 Episode_Reward/pair_body_intrusion: -0.3609 Curriculum/terrain_level_mean: +1.9680 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4759 Episode_Termination/fell_over: 0.0759 Episode_Termination/course_exit: 0.4345 Episode_Termination/illegal_nonwheel_contact: 0.0621 ============================================================================== ============================================================================== Learning iteration 9950/50000 Total steps: 40,755,200 Steps per second: 9,760 Iteration time: 0.420s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0843 Mean episode length: 1279.7 Actor loss: -9.3725 Critic loss: +0.0040 World model loss: +0.9693 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.2565 Episode_Reward/track_angular_velocity: +48.4099 Episode_Reward/upright: +24.8830 Episode_Reward/leg_posture: -2.7195 Episode_Reward/standstill: -6.7445 Episode_Reward/action_rate_l2: -0.5484 Episode_Reward/joint_torques_l2: -0.1506 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0350 Episode_Reward/unnecessary_wheel_lift: -0.1341 Episode_Reward/rolling_leg_motion: -0.7018 Episode_Reward/obstacle_progress: +2.0733 Episode_Reward/wheel_body_intrusion: -0.3321 Episode_Reward/pair_body_intrusion: -0.3713 Curriculum/terrain_level_mean: +1.9722 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1036 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4508 Episode_Termination/fell_over: 0.0984 Episode_Termination/course_exit: 0.4344 Episode_Termination/illegal_nonwheel_contact: 0.0738 ============================================================================== ============================================================================== Learning iteration 10000/50000 Total steps: 40,960,000 Steps per second: 9,779 Iteration time: 0.419s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0845 Mean episode length: 1232.7 Actor loss: -9.4037 Critic loss: +0.0038 World model loss: +0.9849 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +42.6498 Episode_Reward/track_angular_velocity: +46.7028 Episode_Reward/upright: +23.9762 Episode_Reward/leg_posture: -2.5973 Episode_Reward/standstill: -6.3892 Episode_Reward/action_rate_l2: -0.5218 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0265 Episode_Reward/unnecessary_wheel_lift: -0.1113 Episode_Reward/rolling_leg_motion: -0.6803 Episode_Reward/obstacle_progress: +2.0254 Episode_Reward/wheel_body_intrusion: -0.3251 Episode_Reward/pair_body_intrusion: -0.3457 Curriculum/terrain_level_mean: +1.9766 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4211 Episode_Termination/fell_over: 0.0677 Episode_Termination/course_exit: 0.5038 Episode_Termination/illegal_nonwheel_contact: 0.0451 ============================================================================== ============================================================================== Learning iteration 10050/50000 Total steps: 41,164,800 Steps per second: 9,797 Iteration time: 0.418s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0844 Mean episode length: 1336.8 Actor loss: -9.3816 Critic loss: +0.0075 World model loss: +0.9884 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3089 Episode_Reward/track_angular_velocity: +50.5776 Episode_Reward/upright: +25.9871 Episode_Reward/leg_posture: -2.8272 Episode_Reward/standstill: -7.1395 Episode_Reward/action_rate_l2: -0.5624 Episode_Reward/joint_torques_l2: -0.1576 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0287 Episode_Reward/unnecessary_wheel_lift: -0.1341 Episode_Reward/rolling_leg_motion: -0.7426 Episode_Reward/obstacle_progress: +2.2147 Episode_Reward/wheel_body_intrusion: -0.3768 Episode_Reward/pair_body_intrusion: -0.3362 Curriculum/terrain_level_mean: +1.9854 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4785 Episode_Termination/fell_over: 0.0491 Episode_Termination/course_exit: 0.4540 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 10100/50000 Total steps: 41,369,600 Steps per second: 9,815 Iteration time: 0.417s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0845 Mean episode length: 1331.9 Actor loss: -9.3451 Critic loss: +0.0119 World model loss: +1.0160 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9633 Episode_Reward/track_angular_velocity: +50.3619 Episode_Reward/upright: +25.8921 Episode_Reward/leg_posture: -2.7807 Episode_Reward/standstill: -6.7766 Episode_Reward/action_rate_l2: -0.5730 Episode_Reward/joint_torques_l2: -0.1563 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.1294 Episode_Reward/rolling_leg_motion: -0.7469 Episode_Reward/obstacle_progress: +2.1802 Episode_Reward/wheel_body_intrusion: -0.3420 Episode_Reward/pair_body_intrusion: -0.3290 Curriculum/terrain_level_mean: +1.9927 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1045 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4795 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4795 Episode_Termination/illegal_nonwheel_contact: 0.0205 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[454] ============================================================================== Learning iteration 10150/50000 Total steps: 41,574,400 Steps per second: 9,833 Iteration time: 0.417s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0844 Mean episode length: 1315.2 Actor loss: -9.3668 Critic loss: +0.0119 World model loss: +0.9983 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.3414 Episode_Reward/track_angular_velocity: +49.7560 Episode_Reward/upright: +25.5824 Episode_Reward/leg_posture: -2.7681 Episode_Reward/standstill: -6.7917 Episode_Reward/action_rate_l2: -0.5647 Episode_Reward/joint_torques_l2: -0.1530 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0360 Episode_Reward/unnecessary_wheel_lift: -0.1164 Episode_Reward/rolling_leg_motion: -0.7384 Episode_Reward/obstacle_progress: +2.1533 Episode_Reward/wheel_body_intrusion: -0.3403 Episode_Reward/pair_body_intrusion: -0.3174 Curriculum/terrain_level_mean: +1.9978 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4744 Episode_Termination/fell_over: 0.0705 Episode_Termination/course_exit: 0.4359 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== ============================================================================== Learning iteration 10200/50000 Total steps: 41,779,200 Steps per second: 9,851 Iteration time: 0.416s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0838 Mean episode length: 1378.1 Actor loss: -9.3873 Critic loss: +0.0107 World model loss: +1.0640 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4710 Episode_Reward/track_angular_velocity: +52.1568 Episode_Reward/upright: +26.7887 Episode_Reward/leg_posture: -2.9051 Episode_Reward/standstill: -7.7949 Episode_Reward/action_rate_l2: -0.5926 Episode_Reward/joint_torques_l2: -0.1593 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0323 Episode_Reward/unnecessary_wheel_lift: -0.1283 Episode_Reward/rolling_leg_motion: -0.7817 Episode_Reward/obstacle_progress: +2.2116 Episode_Reward/wheel_body_intrusion: -0.3636 Episode_Reward/pair_body_intrusion: -0.3437 Curriculum/terrain_level_mean: +2.0017 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4672 Episode_Termination/fell_over: 0.0511 Episode_Termination/course_exit: 0.4453 Episode_Termination/illegal_nonwheel_contact: 0.0584 ============================================================================== ============================================================================== Learning iteration 10250/50000 Total steps: 41,984,000 Steps per second: 9,869 Iteration time: 0.415s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0837 Mean episode length: 1370.3 Actor loss: -9.3957 Critic loss: +0.0049 World model loss: +0.9794 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1152 Episode_Reward/track_angular_velocity: +51.8215 Episode_Reward/upright: +26.6120 Episode_Reward/leg_posture: -2.9074 Episode_Reward/standstill: -7.7818 Episode_Reward/action_rate_l2: -0.5820 Episode_Reward/joint_torques_l2: -0.1599 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.1312 Episode_Reward/rolling_leg_motion: -0.7623 Episode_Reward/obstacle_progress: +2.2274 Episode_Reward/wheel_body_intrusion: -0.3698 Episode_Reward/pair_body_intrusion: -0.3394 Curriculum/terrain_level_mean: +2.0081 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1060 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4882 Episode_Termination/fell_over: 0.0472 Episode_Termination/course_exit: 0.4409 Episode_Termination/illegal_nonwheel_contact: 0.0394 ============================================================================== ============================================================================== Learning iteration 10300/50000 Total steps: 42,188,800 Steps per second: 9,887 Iteration time: 0.414s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0841 Mean episode length: 1319.8 Actor loss: -9.3765 Critic loss: +0.0046 World model loss: +1.0020 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5266 Episode_Reward/track_angular_velocity: +49.8795 Episode_Reward/upright: +25.6482 Episode_Reward/leg_posture: -2.7460 Episode_Reward/standstill: -7.2633 Episode_Reward/action_rate_l2: -0.5677 Episode_Reward/joint_torques_l2: -0.1529 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0387 Episode_Reward/unnecessary_wheel_lift: -0.1320 Episode_Reward/rolling_leg_motion: -0.7380 Episode_Reward/obstacle_progress: +2.2205 Episode_Reward/wheel_body_intrusion: -0.3387 Episode_Reward/pair_body_intrusion: -0.3617 Curriculum/terrain_level_mean: +2.0125 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1044 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4691 Episode_Termination/fell_over: 0.0617 Episode_Termination/course_exit: 0.4383 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== ============================================================================== Learning iteration 10350/50000 Total steps: 42,393,600 Steps per second: 9,904 Iteration time: 0.414s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0845 Mean episode length: 1334.2 Actor loss: -9.3700 Critic loss: +0.0055 World model loss: +0.9539 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9912 Episode_Reward/track_angular_velocity: +50.4667 Episode_Reward/upright: +25.9199 Episode_Reward/leg_posture: -2.7695 Episode_Reward/standstill: -6.8055 Episode_Reward/action_rate_l2: -0.5725 Episode_Reward/joint_torques_l2: -0.1545 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0182 Episode_Reward/unnecessary_wheel_lift: -0.1028 Episode_Reward/rolling_leg_motion: -0.7509 Episode_Reward/obstacle_progress: +2.1630 Episode_Reward/wheel_body_intrusion: -0.3439 Episode_Reward/pair_body_intrusion: -0.3241 Curriculum/terrain_level_mean: +2.0195 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4891 Episode_Termination/fell_over: 0.0438 Episode_Termination/course_exit: 0.4599 Episode_Termination/illegal_nonwheel_contact: 0.0146 ============================================================================== ============================================================================== Learning iteration 10400/50000 Total steps: 42,598,400 Steps per second: 9,922 Iteration time: 0.413s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0849 Mean episode length: 1400.6 Actor loss: -9.3966 Critic loss: +0.0044 World model loss: +0.9786 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4196 Episode_Reward/track_angular_velocity: +52.9693 Episode_Reward/upright: +27.2176 Episode_Reward/leg_posture: -2.8701 Episode_Reward/standstill: -6.7572 Episode_Reward/action_rate_l2: -0.5961 Episode_Reward/joint_torques_l2: -0.1601 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0274 Episode_Reward/unnecessary_wheel_lift: -0.1200 Episode_Reward/rolling_leg_motion: -0.7986 Episode_Reward/obstacle_progress: +2.2869 Episode_Reward/wheel_body_intrusion: -0.3418 Episode_Reward/pair_body_intrusion: -0.3664 Curriculum/terrain_level_mean: +2.0254 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5430 Episode_Termination/fell_over: 0.0397 Episode_Termination/course_exit: 0.3907 Episode_Termination/illegal_nonwheel_contact: 0.0464 ============================================================================== ============================================================================== Learning iteration 10450/50000 Total steps: 42,803,200 Steps per second: 9,940 Iteration time: 0.412s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0848 Mean episode length: 1366.8 Actor loss: -9.3800 Critic loss: +0.0056 World model loss: +0.9776 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2453 Episode_Reward/track_angular_velocity: +51.6920 Episode_Reward/upright: +26.5830 Episode_Reward/leg_posture: -2.8842 Episode_Reward/standstill: -6.6049 Episode_Reward/action_rate_l2: -0.5759 Episode_Reward/joint_torques_l2: -0.1569 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0347 Episode_Reward/unnecessary_wheel_lift: -0.1336 Episode_Reward/rolling_leg_motion: -0.7779 Episode_Reward/obstacle_progress: +2.2081 Episode_Reward/wheel_body_intrusion: -0.3312 Episode_Reward/pair_body_intrusion: -0.3481 Curriculum/terrain_level_mean: +2.0295 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5325 Episode_Termination/fell_over: 0.0390 Episode_Termination/course_exit: 0.3831 Episode_Termination/illegal_nonwheel_contact: 0.0519 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[995] ============================================================================== Learning iteration 10500/50000 Total steps: 43,008,000 Steps per second: 9,957 Iteration time: 0.411s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0848 Mean episode length: 1390.3 Actor loss: -9.3545 Critic loss: +0.0055 World model loss: +1.0401 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1200 Episode_Reward/track_angular_velocity: +52.6005 Episode_Reward/upright: +27.0231 Episode_Reward/leg_posture: -2.9791 Episode_Reward/standstill: -6.7782 Episode_Reward/action_rate_l2: -0.5775 Episode_Reward/joint_torques_l2: -0.1599 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.1213 Episode_Reward/rolling_leg_motion: -0.7673 Episode_Reward/obstacle_progress: +2.2908 Episode_Reward/wheel_body_intrusion: -0.3871 Episode_Reward/pair_body_intrusion: -0.3795 Curriculum/terrain_level_mean: +2.0366 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1026 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.5704 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.4014 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== ============================================================================== Learning iteration 10550/50000 Total steps: 43,212,800 Steps per second: 9,974 Iteration time: 0.411s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1340.8 Actor loss: -9.3774 Critic loss: +0.0042 World model loss: +0.9005 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2916 Episode_Reward/track_angular_velocity: +50.7377 Episode_Reward/upright: +26.0610 Episode_Reward/leg_posture: -2.8969 Episode_Reward/standstill: -6.5181 Episode_Reward/action_rate_l2: -0.5527 Episode_Reward/joint_torques_l2: -0.1527 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1262 Episode_Reward/rolling_leg_motion: -0.7276 Episode_Reward/obstacle_progress: +2.1026 Episode_Reward/wheel_body_intrusion: -0.3508 Episode_Reward/pair_body_intrusion: -0.3557 Curriculum/terrain_level_mean: +2.0435 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5109 Episode_Termination/fell_over: 0.0292 Episode_Termination/course_exit: 0.4380 Episode_Termination/illegal_nonwheel_contact: 0.0365 ============================================================================== ============================================================================== Learning iteration 10600/50000 Total steps: 43,417,600 Steps per second: 9,992 Iteration time: 0.410s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1352.2 Actor loss: -9.3787 Critic loss: +0.0039 World model loss: +0.9334 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7115 Episode_Reward/track_angular_velocity: +51.2124 Episode_Reward/upright: +26.3160 Episode_Reward/leg_posture: -2.8681 Episode_Reward/standstill: -6.3542 Episode_Reward/action_rate_l2: -0.5636 Episode_Reward/joint_torques_l2: -0.1525 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1138 Episode_Reward/rolling_leg_motion: -0.7511 Episode_Reward/obstacle_progress: +2.0903 Episode_Reward/wheel_body_intrusion: -0.3554 Episode_Reward/pair_body_intrusion: -0.3349 Curriculum/terrain_level_mean: +2.0496 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4898 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4694 Episode_Termination/illegal_nonwheel_contact: 0.0272 ============================================================================== ============================================================================== Learning iteration 10650/50000 Total steps: 43,622,400 Steps per second: 10,009 Iteration time: 0.409s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1419.4 Actor loss: -9.3725 Critic loss: +0.0034 World model loss: +0.8994 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9512 Episode_Reward/track_angular_velocity: +53.7284 Episode_Reward/upright: +27.6479 Episode_Reward/leg_posture: -3.0500 Episode_Reward/standstill: -6.7724 Episode_Reward/action_rate_l2: -0.5955 Episode_Reward/joint_torques_l2: -0.1598 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.1563 Episode_Reward/rolling_leg_motion: -0.8128 Episode_Reward/obstacle_progress: +2.1323 Episode_Reward/wheel_body_intrusion: -0.3757 Episode_Reward/pair_body_intrusion: -0.3682 Curriculum/terrain_level_mean: +2.0579 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1040 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5696 Episode_Termination/fell_over: 0.0127 Episode_Termination/course_exit: 0.3924 Episode_Termination/illegal_nonwheel_contact: 0.0316 ============================================================================== ============================================================================== Learning iteration 10700/50000 Total steps: 43,827,200 Steps per second: 10,026 Iteration time: 0.409s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0845 Mean episode length: 1372.6 Actor loss: -9.3675 Critic loss: +0.0074 World model loss: +0.9739 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3377 Episode_Reward/track_angular_velocity: +51.9752 Episode_Reward/upright: +26.7189 Episode_Reward/leg_posture: -2.9493 Episode_Reward/standstill: -6.9272 Episode_Reward/action_rate_l2: -0.5787 Episode_Reward/joint_torques_l2: -0.1523 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.1284 Episode_Reward/rolling_leg_motion: -0.7805 Episode_Reward/obstacle_progress: +2.1431 Episode_Reward/wheel_body_intrusion: -0.3688 Episode_Reward/pair_body_intrusion: -0.3241 Curriculum/terrain_level_mean: +2.0642 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4855 Episode_Termination/fell_over: 0.0217 Episode_Termination/course_exit: 0.4710 Episode_Termination/illegal_nonwheel_contact: 0.0362 ============================================================================== ============================================================================== Learning iteration 10750/50000 Total steps: 44,032,000 Steps per second: 10,042 Iteration time: 0.408s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0844 Mean episode length: 1406.5 Actor loss: -9.3716 Critic loss: +0.0037 World model loss: +0.9159 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3879 Episode_Reward/track_angular_velocity: +53.3059 Episode_Reward/upright: +27.3724 Episode_Reward/leg_posture: -3.0269 Episode_Reward/standstill: -7.0545 Episode_Reward/action_rate_l2: -0.5858 Episode_Reward/joint_torques_l2: -0.1549 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0110 Episode_Reward/unnecessary_wheel_lift: -0.1376 Episode_Reward/rolling_leg_motion: -0.7832 Episode_Reward/obstacle_progress: +2.1015 Episode_Reward/wheel_body_intrusion: -0.3649 Episode_Reward/pair_body_intrusion: -0.3904 Curriculum/terrain_level_mean: +2.0725 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1045 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.5686 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.4052 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 10800/50000 Total steps: 44,236,800 Steps per second: 10,059 Iteration time: 0.407s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0847 Mean episode length: 1404.1 Actor loss: -9.3604 Critic loss: +0.0036 World model loss: +0.9250 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4329 Episode_Reward/track_angular_velocity: +53.1765 Episode_Reward/upright: +27.3414 Episode_Reward/leg_posture: -3.0202 Episode_Reward/standstill: -6.7479 Episode_Reward/action_rate_l2: -0.5904 Episode_Reward/joint_torques_l2: -0.1553 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0247 Episode_Reward/unnecessary_wheel_lift: -0.1314 Episode_Reward/rolling_leg_motion: -0.7961 Episode_Reward/obstacle_progress: +2.1275 Episode_Reward/wheel_body_intrusion: -0.3542 Episode_Reward/pair_body_intrusion: -0.3805 Curriculum/terrain_level_mean: +2.0798 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1034 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5652 Episode_Termination/fell_over: 0.0217 Episode_Termination/course_exit: 0.3859 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_010800.mp4 ============================================================================== Learning iteration 10850/50000 Total steps: 44,441,600 Steps per second: 10,062 Iteration time: 0.407s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0843 Mean episode length: 1369.1 Actor loss: -9.3987 Critic loss: +0.0025 World model loss: +0.9454 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1331 Episode_Reward/track_angular_velocity: +51.8014 Episode_Reward/upright: +26.6313 Episode_Reward/leg_posture: -2.9919 Episode_Reward/standstill: -6.8831 Episode_Reward/action_rate_l2: -0.5705 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.1594 Episode_Reward/rolling_leg_motion: -0.7553 Episode_Reward/obstacle_progress: +2.1034 Episode_Reward/wheel_body_intrusion: -0.3839 Episode_Reward/pair_body_intrusion: -0.3881 Curriculum/terrain_level_mean: +2.0891 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.5032 Episode_Termination/fell_over: 0.0064 Episode_Termination/course_exit: 0.4713 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 10900/50000 Total steps: 44,646,400 Steps per second: 10,079 Iteration time: 0.406s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0847 Mean episode length: 1392.0 Actor loss: -9.3910 Critic loss: +0.0032 World model loss: +0.9059 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9544 Episode_Reward/track_angular_velocity: +52.6642 Episode_Reward/upright: +27.0727 Episode_Reward/leg_posture: -3.0184 Episode_Reward/standstill: -6.5802 Episode_Reward/action_rate_l2: -0.5950 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0245 Episode_Reward/unnecessary_wheel_lift: -0.1313 Episode_Reward/rolling_leg_motion: -0.7699 Episode_Reward/obstacle_progress: +2.1699 Episode_Reward/wheel_body_intrusion: -0.3696 Episode_Reward/pair_body_intrusion: -0.3800 Curriculum/terrain_level_mean: +2.0972 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1040 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5767 Episode_Termination/fell_over: 0.0307 Episode_Termination/course_exit: 0.3681 Episode_Termination/illegal_nonwheel_contact: 0.0429 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1458] ============================================================================== Learning iteration 10950/50000 Total steps: 44,851,200 Steps per second: 10,095 Iteration time: 0.406s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0845 Mean episode length: 1426.0 Actor loss: -9.3527 Critic loss: +0.0062 World model loss: +0.9322 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0387 Episode_Reward/track_angular_velocity: +53.9694 Episode_Reward/upright: +27.7591 Episode_Reward/leg_posture: -3.0865 Episode_Reward/standstill: -6.7122 Episode_Reward/action_rate_l2: -0.6111 Episode_Reward/joint_torques_l2: -0.1552 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0306 Episode_Reward/unnecessary_wheel_lift: -0.1609 Episode_Reward/rolling_leg_motion: -0.8113 Episode_Reward/obstacle_progress: +2.1586 Episode_Reward/wheel_body_intrusion: -0.3702 Episode_Reward/pair_body_intrusion: -0.4281 Curriculum/terrain_level_mean: +2.1025 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5928 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.3413 Episode_Termination/illegal_nonwheel_contact: 0.0359 ============================================================================== ============================================================================== Learning iteration 11000/50000 Total steps: 45,056,000 Steps per second: 10,111 Iteration time: 0.405s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0846 Mean episode length: 1381.5 Actor loss: -9.3594 Critic loss: +0.0045 World model loss: +0.9513 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3993 Episode_Reward/track_angular_velocity: +52.2366 Episode_Reward/upright: +26.8739 Episode_Reward/leg_posture: -3.0053 Episode_Reward/standstill: -6.3622 Episode_Reward/action_rate_l2: -0.5971 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0270 Episode_Reward/unnecessary_wheel_lift: -0.1375 Episode_Reward/rolling_leg_motion: -0.7741 Episode_Reward/obstacle_progress: +2.1547 Episode_Reward/wheel_body_intrusion: -0.3703 Episode_Reward/pair_body_intrusion: -0.3559 Curriculum/terrain_level_mean: +2.1082 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1060 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.5602 Episode_Termination/fell_over: 0.0120 Episode_Termination/course_exit: 0.3795 Episode_Termination/illegal_nonwheel_contact: 0.0602 ============================================================================== ============================================================================== Learning iteration 11050/50000 Total steps: 45,260,800 Steps per second: 10,127 Iteration time: 0.404s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1381.2 Actor loss: -9.3391 Critic loss: +0.0111 World model loss: +0.9378 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5097 Episode_Reward/track_angular_velocity: +52.2984 Episode_Reward/upright: +26.9009 Episode_Reward/leg_posture: -2.9781 Episode_Reward/standstill: -6.6133 Episode_Reward/action_rate_l2: -0.5923 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0216 Episode_Reward/unnecessary_wheel_lift: -0.1325 Episode_Reward/rolling_leg_motion: -0.7755 Episode_Reward/obstacle_progress: +2.1336 Episode_Reward/wheel_body_intrusion: -0.3441 Episode_Reward/pair_body_intrusion: -0.3363 Curriculum/terrain_level_mean: +2.1157 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1046 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5032 Episode_Termination/fell_over: 0.0387 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[912] ============================================================================== Learning iteration 11100/50000 Total steps: 45,465,600 Steps per second: 10,143 Iteration time: 0.404s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1412.9 Actor loss: -9.3631 Critic loss: +0.0065 World model loss: +0.9003 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7340 Episode_Reward/track_angular_velocity: +53.5773 Episode_Reward/upright: +27.5442 Episode_Reward/leg_posture: -3.0176 Episode_Reward/standstill: -6.7186 Episode_Reward/action_rate_l2: -0.5864 Episode_Reward/joint_torques_l2: -0.1548 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.1292 Episode_Reward/rolling_leg_motion: -0.7853 Episode_Reward/obstacle_progress: +2.1868 Episode_Reward/wheel_body_intrusion: -0.3505 Episode_Reward/pair_body_intrusion: -0.3536 Curriculum/terrain_level_mean: +2.1223 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.6056 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.3521 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1249] ============================================================================== Learning iteration 11150/50000 Total steps: 45,670,400 Steps per second: 10,160 Iteration time: 0.403s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1437.7 Actor loss: -9.3994 Critic loss: +0.0029 World model loss: +0.9451 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4577 Episode_Reward/track_angular_velocity: +54.4747 Episode_Reward/upright: +28.0078 Episode_Reward/leg_posture: -3.0512 Episode_Reward/standstill: -6.7891 Episode_Reward/action_rate_l2: -0.5981 Episode_Reward/joint_torques_l2: -0.1558 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0225 Episode_Reward/unnecessary_wheel_lift: -0.1417 Episode_Reward/rolling_leg_motion: -0.7776 Episode_Reward/obstacle_progress: +2.3284 Episode_Reward/wheel_body_intrusion: -0.3737 Episode_Reward/pair_body_intrusion: -0.4180 Curriculum/terrain_level_mean: +2.1292 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.5732 Episode_Termination/fell_over: 0.0305 Episode_Termination/course_exit: 0.3780 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2041] ============================================================================== Learning iteration 11200/50000 Total steps: 45,875,200 Steps per second: 10,176 Iteration time: 0.403s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0844 Mean episode length: 1352.3 Actor loss: -9.3599 Critic loss: +0.0068 World model loss: +0.8684 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3593 Episode_Reward/track_angular_velocity: +51.1670 Episode_Reward/upright: +26.3115 Episode_Reward/leg_posture: -2.9210 Episode_Reward/standstill: -6.6319 Episode_Reward/action_rate_l2: -0.5605 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0264 Episode_Reward/unnecessary_wheel_lift: -0.1198 Episode_Reward/rolling_leg_motion: -0.7214 Episode_Reward/obstacle_progress: +2.1682 Episode_Reward/wheel_body_intrusion: -0.3672 Episode_Reward/pair_body_intrusion: -0.3451 Curriculum/terrain_level_mean: +2.1355 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1058 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5188 Episode_Termination/fell_over: 0.0226 Episode_Termination/course_exit: 0.4060 Episode_Termination/illegal_nonwheel_contact: 0.0677 ============================================================================== ============================================================================== Learning iteration 11250/50000 Total steps: 46,080,000 Steps per second: 10,192 Iteration time: 0.402s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1381.6 Actor loss: -9.3545 Critic loss: +0.0050 World model loss: +0.9202 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4288 Episode_Reward/track_angular_velocity: +52.3478 Episode_Reward/upright: +26.9145 Episode_Reward/leg_posture: -2.9586 Episode_Reward/standstill: -6.5637 Episode_Reward/action_rate_l2: -0.5791 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.1235 Episode_Reward/rolling_leg_motion: -0.7566 Episode_Reward/obstacle_progress: +2.1872 Episode_Reward/wheel_body_intrusion: -0.3595 Episode_Reward/pair_body_intrusion: -0.3441 Curriculum/terrain_level_mean: +2.1423 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.5833 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.3846 Episode_Termination/illegal_nonwheel_contact: 0.0321 ============================================================================== ============================================================================== Learning iteration 11300/50000 Total steps: 46,284,800 Steps per second: 10,207 Iteration time: 0.401s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1391.8 Actor loss: -9.3347 Critic loss: +0.0114 World model loss: +0.8939 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8344 Episode_Reward/track_angular_velocity: +52.7191 Episode_Reward/upright: +27.1024 Episode_Reward/leg_posture: -2.9869 Episode_Reward/standstill: -6.5185 Episode_Reward/action_rate_l2: -0.5949 Episode_Reward/joint_torques_l2: -0.1499 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0299 Episode_Reward/unnecessary_wheel_lift: -0.1517 Episode_Reward/rolling_leg_motion: -0.7527 Episode_Reward/obstacle_progress: +2.2240 Episode_Reward/wheel_body_intrusion: -0.3623 Episode_Reward/pair_body_intrusion: -0.3535 Curriculum/terrain_level_mean: +2.1487 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5423 Episode_Termination/fell_over: 0.0493 Episode_Termination/course_exit: 0.3803 Episode_Termination/illegal_nonwheel_contact: 0.0563 ============================================================================== ============================================================================== Learning iteration 11350/50000 Total steps: 46,489,600 Steps per second: 10,223 Iteration time: 0.401s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1336.3 Actor loss: -9.3793 Critic loss: +0.0032 World model loss: +0.9448 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9730 Episode_Reward/track_angular_velocity: +50.6265 Episode_Reward/upright: +26.0178 Episode_Reward/leg_posture: -2.8662 Episode_Reward/standstill: -6.2880 Episode_Reward/action_rate_l2: -0.5639 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0444 Episode_Reward/unnecessary_wheel_lift: -0.1276 Episode_Reward/rolling_leg_motion: -0.7235 Episode_Reward/obstacle_progress: +2.1515 Episode_Reward/wheel_body_intrusion: -0.3561 Episode_Reward/pair_body_intrusion: -0.3145 Curriculum/terrain_level_mean: +2.1519 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1043 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.5141 Episode_Termination/fell_over: 0.0775 Episode_Termination/course_exit: 0.3662 Episode_Termination/illegal_nonwheel_contact: 0.0915 ============================================================================== ============================================================================== Learning iteration 11400/50000 Total steps: 46,694,400 Steps per second: 10,239 Iteration time: 0.400s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0848 Mean episode length: 1363.2 Actor loss: -9.3188 Critic loss: +0.0038 World model loss: +0.8620 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9411 Episode_Reward/track_angular_velocity: +51.5377 Episode_Reward/upright: +26.5278 Episode_Reward/leg_posture: -2.9602 Episode_Reward/standstill: -6.2740 Episode_Reward/action_rate_l2: -0.5835 Episode_Reward/joint_torques_l2: -0.1497 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0240 Episode_Reward/unnecessary_wheel_lift: -0.1078 Episode_Reward/rolling_leg_motion: -0.7362 Episode_Reward/obstacle_progress: +2.2569 Episode_Reward/wheel_body_intrusion: -0.3937 Episode_Reward/pair_body_intrusion: -0.3915 Curriculum/terrain_level_mean: +2.1584 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1041 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4783 Episode_Termination/fell_over: 0.0522 Episode_Termination/course_exit: 0.4435 Episode_Termination/illegal_nonwheel_contact: 0.0609 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[574] ============================================================================== Learning iteration 11450/50000 Total steps: 46,899,200 Steps per second: 10,255 Iteration time: 0.399s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0849 Mean episode length: 1380.7 Actor loss: -9.3627 Critic loss: +0.0040 World model loss: +0.9197 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3996 Episode_Reward/track_angular_velocity: +52.1971 Episode_Reward/upright: +26.8607 Episode_Reward/leg_posture: -2.9758 Episode_Reward/standstill: -6.0396 Episode_Reward/action_rate_l2: -0.5821 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1258 Episode_Reward/rolling_leg_motion: -0.7539 Episode_Reward/obstacle_progress: +2.2021 Episode_Reward/wheel_body_intrusion: -0.3557 Episode_Reward/pair_body_intrusion: -0.4335 Curriculum/terrain_level_mean: +2.1650 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1051 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.5328 Episode_Termination/fell_over: 0.0438 Episode_Termination/course_exit: 0.4161 Episode_Termination/illegal_nonwheel_contact: 0.0292 ============================================================================== ============================================================================== Learning iteration 11500/50000 Total steps: 47,104,000 Steps per second: 10,271 Iteration time: 0.399s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0848 Mean episode length: 1405.3 Actor loss: -9.3703 Critic loss: +0.0038 World model loss: +0.8988 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2990 Episode_Reward/track_angular_velocity: +53.1665 Episode_Reward/upright: +27.3629 Episode_Reward/leg_posture: -3.0220 Episode_Reward/standstill: -6.3636 Episode_Reward/action_rate_l2: -0.6040 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0274 Episode_Reward/unnecessary_wheel_lift: -0.1328 Episode_Reward/rolling_leg_motion: -0.7826 Episode_Reward/obstacle_progress: +2.2212 Episode_Reward/wheel_body_intrusion: -0.3715 Episode_Reward/pair_body_intrusion: -0.4020 Curriculum/terrain_level_mean: +2.1709 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1047 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.5921 Episode_Termination/fell_over: 0.0395 Episode_Termination/course_exit: 0.3421 Episode_Termination/illegal_nonwheel_contact: 0.0461 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1351] ============================================================================== Learning iteration 11550/50000 Total steps: 47,308,800 Steps per second: 10,285 Iteration time: 0.398s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0846 Mean episode length: 1321.5 Actor loss: -9.3380 Critic loss: +0.0059 World model loss: +0.9160 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.4946 Episode_Reward/track_angular_velocity: +49.9685 Episode_Reward/upright: +25.7259 Episode_Reward/leg_posture: -2.8658 Episode_Reward/standstill: -6.2846 Episode_Reward/action_rate_l2: -0.5655 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.1279 Episode_Reward/rolling_leg_motion: -0.7306 Episode_Reward/obstacle_progress: +2.1702 Episode_Reward/wheel_body_intrusion: -0.3760 Episode_Reward/pair_body_intrusion: -0.3839 Curriculum/terrain_level_mean: +2.1763 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1042 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4646 Episode_Termination/fell_over: 0.0236 Episode_Termination/course_exit: 0.4803 Episode_Termination/illegal_nonwheel_contact: 0.0551 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 11600/50000 Total steps: 47,513,600 Steps per second: 10,301 Iteration time: 0.398s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0851 Mean episode length: 1321.8 Actor loss: -9.3266 Critic loss: +0.0045 World model loss: +0.8769 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6230 Episode_Reward/track_angular_velocity: +50.0812 Episode_Reward/upright: +25.7494 Episode_Reward/leg_posture: -2.8166 Episode_Reward/standstill: -6.0006 Episode_Reward/action_rate_l2: -0.5436 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.1127 Episode_Reward/rolling_leg_motion: -0.7058 Episode_Reward/obstacle_progress: +2.1525 Episode_Reward/wheel_body_intrusion: -0.3595 Episode_Reward/pair_body_intrusion: -0.3998 Curriculum/terrain_level_mean: +2.1846 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1029 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4820 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.4676 Episode_Termination/illegal_nonwheel_contact: 0.0288 ============================================================================== ============================================================================== Learning iteration 11650/50000 Total steps: 47,718,400 Steps per second: 10,316 Iteration time: 0.397s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1330.0 Actor loss: -9.3858 Critic loss: +0.0094 World model loss: +0.8833 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7746 Episode_Reward/track_angular_velocity: +50.2752 Episode_Reward/upright: +25.8883 Episode_Reward/leg_posture: -2.8687 Episode_Reward/standstill: -6.2064 Episode_Reward/action_rate_l2: -0.5586 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.1348 Episode_Reward/rolling_leg_motion: -0.7334 Episode_Reward/obstacle_progress: +2.1767 Episode_Reward/wheel_body_intrusion: -0.3620 Episode_Reward/pair_body_intrusion: -0.3663 Curriculum/terrain_level_mean: +2.1895 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1041 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5229 Episode_Termination/fell_over: 0.0458 Episode_Termination/course_exit: 0.4118 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 11700/50000 Total steps: 47,923,200 Steps per second: 10,331 Iteration time: 0.396s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0851 Mean episode length: 1365.6 Actor loss: -9.3588 Critic loss: +0.0042 World model loss: +0.9069 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2239 Episode_Reward/track_angular_velocity: +51.6570 Episode_Reward/upright: +26.5943 Episode_Reward/leg_posture: -2.9168 Episode_Reward/standstill: -6.2278 Episode_Reward/action_rate_l2: -0.5819 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.1296 Episode_Reward/rolling_leg_motion: -0.7434 Episode_Reward/obstacle_progress: +2.3203 Episode_Reward/wheel_body_intrusion: -0.3962 Episode_Reward/pair_body_intrusion: -0.4037 Curriculum/terrain_level_mean: +2.1948 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5407 Episode_Termination/fell_over: 0.0222 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0444 ============================================================================== ============================================================================== Learning iteration 11750/50000 Total steps: 48,128,000 Steps per second: 10,346 Iteration time: 0.396s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1360.7 Actor loss: -9.3629 Critic loss: +0.0026 World model loss: +0.8897 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8846 Episode_Reward/track_angular_velocity: +51.3663 Episode_Reward/upright: +26.5093 Episode_Reward/leg_posture: -2.8847 Episode_Reward/standstill: -6.3451 Episode_Reward/action_rate_l2: -0.5914 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0186 Episode_Reward/unnecessary_wheel_lift: -0.1221 Episode_Reward/rolling_leg_motion: -0.7423 Episode_Reward/obstacle_progress: +2.3533 Episode_Reward/wheel_body_intrusion: -0.3730 Episode_Reward/pair_body_intrusion: -0.4571 Curriculum/terrain_level_mean: +2.2031 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1039 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4965 Episode_Termination/fell_over: 0.0426 Episode_Termination/course_exit: 0.4539 Episode_Termination/illegal_nonwheel_contact: 0.0426 ============================================================================== ============================================================================== Learning iteration 11800/50000 Total steps: 48,332,800 Steps per second: 10,361 Iteration time: 0.395s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1368.9 Actor loss: -9.3672 Critic loss: +0.0036 World model loss: +0.8698 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0358 Episode_Reward/track_angular_velocity: +51.6910 Episode_Reward/upright: +26.6848 Episode_Reward/leg_posture: -2.9005 Episode_Reward/standstill: -6.4159 Episode_Reward/action_rate_l2: -0.5799 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1249 Episode_Reward/rolling_leg_motion: -0.7327 Episode_Reward/obstacle_progress: +2.2812 Episode_Reward/wheel_body_intrusion: -0.3694 Episode_Reward/pair_body_intrusion: -0.4401 Curriculum/terrain_level_mean: +2.2114 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1051 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.5318 Episode_Termination/fell_over: 0.0289 Episode_Termination/course_exit: 0.4220 Episode_Termination/illegal_nonwheel_contact: 0.0231 ============================================================================== ============================================================================== Learning iteration 11850/50000 Total steps: 48,537,600 Steps per second: 10,376 Iteration time: 0.395s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0846 Mean episode length: 1417.4 Actor loss: -9.3389 Critic loss: +0.0057 World model loss: +0.9105 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7598 Episode_Reward/track_angular_velocity: +53.4940 Episode_Reward/upright: +27.6074 Episode_Reward/leg_posture: -3.0146 Episode_Reward/standstill: -6.7032 Episode_Reward/action_rate_l2: -0.6052 Episode_Reward/joint_torques_l2: -0.1562 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.1299 Episode_Reward/rolling_leg_motion: -0.7803 Episode_Reward/obstacle_progress: +2.4061 Episode_Reward/wheel_body_intrusion: -0.4106 Episode_Reward/pair_body_intrusion: -0.4811 Curriculum/terrain_level_mean: +2.2175 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.5473 Episode_Termination/fell_over: 0.0135 Episode_Termination/course_exit: 0.4189 Episode_Termination/illegal_nonwheel_contact: 0.0270 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1771] ============================================================================== Learning iteration 11900/50000 Total steps: 48,742,400 Steps per second: 10,391 Iteration time: 0.394s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0850 Mean episode length: 1369.5 Actor loss: -9.3751 Critic loss: +0.0024 World model loss: +0.8746 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2316 Episode_Reward/track_angular_velocity: +51.6850 Episode_Reward/upright: +26.6809 Episode_Reward/leg_posture: -2.9158 Episode_Reward/standstill: -6.0222 Episode_Reward/action_rate_l2: -0.5788 Episode_Reward/joint_torques_l2: -0.1510 Episode_Reward/dof_pos_limits: -0.0023 Episode_Reward/terminated: -0.0321 Episode_Reward/unnecessary_wheel_lift: -0.1270 Episode_Reward/rolling_leg_motion: -0.7512 Episode_Reward/obstacle_progress: +2.2884 Episode_Reward/wheel_body_intrusion: -0.4183 Episode_Reward/pair_body_intrusion: -0.4295 Curriculum/terrain_level_mean: +2.2224 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1040 Metrics/base_velocity/error_vel_yaw: 0.1378 Episode_Termination/time_out: 0.5190 Episode_Termination/fell_over: 0.0253 Episode_Termination/course_exit: 0.4051 Episode_Termination/illegal_nonwheel_contact: 0.0696 ============================================================================== ============================================================================== Learning iteration 11950/50000 Total steps: 48,947,200 Steps per second: 10,405 Iteration time: 0.394s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0851 Mean episode length: 1336.1 Actor loss: -9.4135 Critic loss: +0.0039 World model loss: +0.8824 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1252 Episode_Reward/track_angular_velocity: +50.4411 Episode_Reward/upright: +26.0316 Episode_Reward/leg_posture: -2.8516 Episode_Reward/standstill: -5.8197 Episode_Reward/action_rate_l2: -0.5691 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0020 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.1206 Episode_Reward/rolling_leg_motion: -0.7438 Episode_Reward/obstacle_progress: +2.2136 Episode_Reward/wheel_body_intrusion: -0.3847 Episode_Reward/pair_body_intrusion: -0.4002 Curriculum/terrain_level_mean: +2.2271 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1034 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5429 Episode_Termination/fell_over: 0.0500 Episode_Termination/course_exit: 0.3714 Episode_Termination/illegal_nonwheel_contact: 0.0643 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[705] ============================================================================== Learning iteration 12000/50000 Total steps: 49,152,000 Steps per second: 10,420 Iteration time: 0.393s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1348.1 Actor loss: -9.3694 Critic loss: +0.0046 World model loss: +0.8584 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4557 Episode_Reward/track_angular_velocity: +50.9330 Episode_Reward/upright: +26.2604 Episode_Reward/leg_posture: -2.8370 Episode_Reward/standstill: -6.3340 Episode_Reward/action_rate_l2: -0.5731 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0027 Episode_Reward/terminated: -0.0237 Episode_Reward/unnecessary_wheel_lift: -0.1150 Episode_Reward/rolling_leg_motion: -0.7346 Episode_Reward/obstacle_progress: +2.2042 Episode_Reward/wheel_body_intrusion: -0.3977 Episode_Reward/pair_body_intrusion: -0.3758 Curriculum/terrain_level_mean: +2.2332 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4586 Episode_Termination/fell_over: 0.0226 Episode_Termination/course_exit: 0.4737 Episode_Termination/illegal_nonwheel_contact: 0.0526 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2419] ============================================================================== Learning iteration 12050/50000 Total steps: 49,356,800 Steps per second: 10,434 Iteration time: 0.393s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0849 Mean episode length: 1440.4 Actor loss: -9.3697 Critic loss: +0.0046 World model loss: +0.8571 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6062 Episode_Reward/track_angular_velocity: +54.3965 Episode_Reward/upright: +28.0245 Episode_Reward/leg_posture: -3.0508 Episode_Reward/standstill: -6.4734 Episode_Reward/action_rate_l2: -0.6147 Episode_Reward/joint_torques_l2: -0.1586 Episode_Reward/dof_pos_limits: -0.0020 Episode_Reward/terminated: -0.0225 Episode_Reward/unnecessary_wheel_lift: -0.1398 Episode_Reward/rolling_leg_motion: -0.7938 Episode_Reward/obstacle_progress: +2.3926 Episode_Reward/wheel_body_intrusion: -0.4133 Episode_Reward/pair_body_intrusion: -0.4166 Curriculum/terrain_level_mean: +2.2390 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1044 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.5677 Episode_Termination/fell_over: 0.0129 Episode_Termination/course_exit: 0.3806 Episode_Termination/illegal_nonwheel_contact: 0.0516 ============================================================================== ============================================================================== Learning iteration 12100/50000 Total steps: 49,561,600 Steps per second: 10,449 Iteration time: 0.392s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0844 Mean episode length: 1356.2 Actor loss: -9.4057 Critic loss: +0.0029 World model loss: +0.8546 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5479 Episode_Reward/track_angular_velocity: +51.1627 Episode_Reward/upright: +26.3633 Episode_Reward/leg_posture: -2.9049 Episode_Reward/standstill: -6.6306 Episode_Reward/action_rate_l2: -0.5878 Episode_Reward/joint_torques_l2: -0.1522 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0291 Episode_Reward/unnecessary_wheel_lift: -0.1557 Episode_Reward/rolling_leg_motion: -0.7532 Episode_Reward/obstacle_progress: +2.3378 Episode_Reward/wheel_body_intrusion: -0.4000 Episode_Reward/pair_body_intrusion: -0.4004 Curriculum/terrain_level_mean: +2.2463 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1061 Metrics/base_velocity/error_vel_yaw: 0.1380 Episode_Termination/time_out: 0.4906 Episode_Termination/fell_over: 0.0377 Episode_Termination/course_exit: 0.4403 Episode_Termination/illegal_nonwheel_contact: 0.0440 ============================================================================== ============================================================================== Learning iteration 12150/50000 Total steps: 49,766,400 Steps per second: 10,464 Iteration time: 0.391s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0846 Mean episode length: 1338.2 Actor loss: -9.3779 Critic loss: +0.0033 World model loss: +0.8748 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0072 Episode_Reward/track_angular_velocity: +50.4923 Episode_Reward/upright: +26.0080 Episode_Reward/leg_posture: -2.8365 Episode_Reward/standstill: -6.2469 Episode_Reward/action_rate_l2: -0.5875 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0366 Episode_Reward/unnecessary_wheel_lift: -0.1566 Episode_Reward/rolling_leg_motion: -0.7341 Episode_Reward/obstacle_progress: +2.2422 Episode_Reward/wheel_body_intrusion: -0.4073 Episode_Reward/pair_body_intrusion: -0.3388 Curriculum/terrain_level_mean: +2.2495 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1052 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.4845 Episode_Termination/fell_over: 0.0311 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0683 ============================================================================== ============================================================================== Learning iteration 12200/50000 Total steps: 49,971,200 Steps per second: 10,478 Iteration time: 0.391s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0846 Mean episode length: 1380.1 Actor loss: -9.4153 Critic loss: +0.0047 World model loss: +0.9150 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6243 Episode_Reward/track_angular_velocity: +52.1066 Episode_Reward/upright: +26.8330 Episode_Reward/leg_posture: -2.9453 Episode_Reward/standstill: -6.7737 Episode_Reward/action_rate_l2: -0.5987 Episode_Reward/joint_torques_l2: -0.1527 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0404 Episode_Reward/unnecessary_wheel_lift: -0.1325 Episode_Reward/rolling_leg_motion: -0.7571 Episode_Reward/obstacle_progress: +2.3593 Episode_Reward/wheel_body_intrusion: -0.4234 Episode_Reward/pair_body_intrusion: -0.3513 Curriculum/terrain_level_mean: +2.2534 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1039 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.5404 Episode_Termination/fell_over: 0.0621 Episode_Termination/course_exit: 0.3665 Episode_Termination/illegal_nonwheel_contact: 0.0621 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1193] ============================================================================== Learning iteration 12250/50000 Total steps: 50,176,000 Steps per second: 10,492 Iteration time: 0.390s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0848 Mean episode length: 1351.8 Actor loss: -9.3994 Critic loss: +0.0038 World model loss: +0.9035 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7259 Episode_Reward/track_angular_velocity: +51.0720 Episode_Reward/upright: +26.3027 Episode_Reward/leg_posture: -2.8947 Episode_Reward/standstill: -6.4329 Episode_Reward/action_rate_l2: -0.5907 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.1409 Episode_Reward/rolling_leg_motion: -0.7376 Episode_Reward/obstacle_progress: +2.3897 Episode_Reward/wheel_body_intrusion: -0.4211 Episode_Reward/pair_body_intrusion: -0.4033 Curriculum/terrain_level_mean: +2.2598 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1029 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0256 Episode_Termination/course_exit: 0.4423 Episode_Termination/illegal_nonwheel_contact: 0.0385 ============================================================================== ============================================================================== Learning iteration 12300/50000 Total steps: 50,380,800 Steps per second: 10,506 Iteration time: 0.390s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0845 Mean episode length: 1384.4 Actor loss: -9.3485 Critic loss: +0.0058 World model loss: +0.8779 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7033 Episode_Reward/track_angular_velocity: +52.2824 Episode_Reward/upright: +26.9493 Episode_Reward/leg_posture: -2.8966 Episode_Reward/standstill: -6.8374 Episode_Reward/action_rate_l2: -0.6128 Episode_Reward/joint_torques_l2: -0.1548 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0324 Episode_Reward/unnecessary_wheel_lift: -0.1568 Episode_Reward/rolling_leg_motion: -0.7605 Episode_Reward/obstacle_progress: +2.3689 Episode_Reward/wheel_body_intrusion: -0.4022 Episode_Reward/pair_body_intrusion: -0.4370 Curriculum/terrain_level_mean: +2.2661 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1041 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.4257 Episode_Termination/illegal_nonwheel_contact: 0.0676 ============================================================================== ============================================================================== Learning iteration 12350/50000 Total steps: 50,585,600 Steps per second: 10,520 Iteration time: 0.389s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0846 Mean episode length: 1385.3 Actor loss: -9.3925 Critic loss: +0.0040 World model loss: +0.8596 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7109 Episode_Reward/track_angular_velocity: +52.1924 Episode_Reward/upright: +26.9279 Episode_Reward/leg_posture: -2.9096 Episode_Reward/standstill: -6.7086 Episode_Reward/action_rate_l2: -0.6143 Episode_Reward/joint_torques_l2: -0.1546 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.1417 Episode_Reward/rolling_leg_motion: -0.7521 Episode_Reward/obstacle_progress: +2.4924 Episode_Reward/wheel_body_intrusion: -0.3862 Episode_Reward/pair_body_intrusion: -0.4257 Curriculum/terrain_level_mean: +2.2712 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1048 Metrics/base_velocity/error_vel_yaw: 0.1394 Episode_Termination/time_out: 0.5400 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.3933 Episode_Termination/illegal_nonwheel_contact: 0.0533 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1800] ============================================================================== Learning iteration 12400/50000 Total steps: 50,790,400 Steps per second: 10,534 Iteration time: 0.389s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0844 Mean episode length: 1404.5 Actor loss: -9.3760 Critic loss: +0.0053 World model loss: +0.8745 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3556 Episode_Reward/track_angular_velocity: +52.9601 Episode_Reward/upright: +27.3038 Episode_Reward/leg_posture: -2.9370 Episode_Reward/standstill: -7.1543 Episode_Reward/action_rate_l2: -0.6167 Episode_Reward/joint_torques_l2: -0.1575 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0274 Episode_Reward/unnecessary_wheel_lift: -0.1321 Episode_Reward/rolling_leg_motion: -0.7543 Episode_Reward/obstacle_progress: +2.4742 Episode_Reward/wheel_body_intrusion: -0.3593 Episode_Reward/pair_body_intrusion: -0.3854 Curriculum/terrain_level_mean: +2.2766 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1049 Metrics/base_velocity/error_vel_yaw: 0.1386 Episode_Termination/time_out: 0.5524 Episode_Termination/fell_over: 0.0350 Episode_Termination/course_exit: 0.3776 Episode_Termination/illegal_nonwheel_contact: 0.0559 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[659] ============================================================================== Learning iteration 12450/50000 Total steps: 50,995,200 Steps per second: 10,548 Iteration time: 0.388s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0846 Mean episode length: 1380.1 Actor loss: -9.3672 Critic loss: +0.0045 World model loss: +0.8959 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6716 Episode_Reward/track_angular_velocity: +52.0972 Episode_Reward/upright: +26.8464 Episode_Reward/leg_posture: -2.9295 Episode_Reward/standstill: -6.9185 Episode_Reward/action_rate_l2: -0.5907 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1186 Episode_Reward/rolling_leg_motion: -0.7262 Episode_Reward/obstacle_progress: +2.4194 Episode_Reward/wheel_body_intrusion: -0.3700 Episode_Reward/pair_body_intrusion: -0.4108 Curriculum/terrain_level_mean: +2.2808 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1037 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.5391 Episode_Termination/fell_over: 0.0348 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0348 ============================================================================== ============================================================================== Learning iteration 12500/50000 Total steps: 51,200,000 Steps per second: 10,561 Iteration time: 0.388s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1373.7 Actor loss: -9.3733 Critic loss: +0.0084 World model loss: +0.9200 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4225 Episode_Reward/track_angular_velocity: +51.8692 Episode_Reward/upright: +26.7157 Episode_Reward/leg_posture: -2.9199 Episode_Reward/standstill: -6.7751 Episode_Reward/action_rate_l2: -0.5776 Episode_Reward/joint_torques_l2: -0.1546 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0402 Episode_Reward/unnecessary_wheel_lift: -0.1264 Episode_Reward/rolling_leg_motion: -0.7399 Episode_Reward/obstacle_progress: +2.4301 Episode_Reward/wheel_body_intrusion: -0.3467 Episode_Reward/pair_body_intrusion: -0.3717 Curriculum/terrain_level_mean: +2.2859 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1038 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.5309 Episode_Termination/fell_over: 0.0370 Episode_Termination/course_exit: 0.3765 Episode_Termination/illegal_nonwheel_contact: 0.0741 ============================================================================== ============================================================================== Learning iteration 12550/50000 Total steps: 51,404,800 Steps per second: 10,575 Iteration time: 0.387s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0849 Mean episode length: 1386.1 Actor loss: -9.3757 Critic loss: +0.0045 World model loss: +0.8781 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9599 Episode_Reward/track_angular_velocity: +52.3800 Episode_Reward/upright: +26.9731 Episode_Reward/leg_posture: -2.9147 Episode_Reward/standstill: -6.7743 Episode_Reward/action_rate_l2: -0.5974 Episode_Reward/joint_torques_l2: -0.1546 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0244 Episode_Reward/unnecessary_wheel_lift: -0.1469 Episode_Reward/rolling_leg_motion: -0.7507 Episode_Reward/obstacle_progress: +2.3732 Episode_Reward/wheel_body_intrusion: -0.3415 Episode_Reward/pair_body_intrusion: -0.3473 Curriculum/terrain_level_mean: +2.2942 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1032 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.5667 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.3833 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== ============================================================================== Learning iteration 12600/50000 Total steps: 51,609,600 Steps per second: 10,589 Iteration time: 0.387s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0846 Mean episode length: 1359.0 Actor loss: -9.3732 Critic loss: +0.0042 World model loss: +0.8789 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0388 Episode_Reward/track_angular_velocity: +51.3488 Episode_Reward/upright: +26.4703 Episode_Reward/leg_posture: -2.8823 Episode_Reward/standstill: -6.9188 Episode_Reward/action_rate_l2: -0.5816 Episode_Reward/joint_torques_l2: -0.1527 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0265 Episode_Reward/unnecessary_wheel_lift: -0.1394 Episode_Reward/rolling_leg_motion: -0.7327 Episode_Reward/obstacle_progress: +2.3061 Episode_Reward/wheel_body_intrusion: -0.3530 Episode_Reward/pair_body_intrusion: -0.3542 Curriculum/terrain_level_mean: +2.2993 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1028 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.5090 Episode_Termination/fell_over: 0.0299 Episode_Termination/course_exit: 0.4311 Episode_Termination/illegal_nonwheel_contact: 0.0539 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[949] ============================================================================== Learning iteration 12650/50000 Total steps: 51,814,400 Steps per second: 10,603 Iteration time: 0.386s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1346.1 Actor loss: -9.4109 Critic loss: +0.0035 World model loss: +0.8820 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6208 Episode_Reward/track_angular_velocity: +50.8147 Episode_Reward/upright: +26.1869 Episode_Reward/leg_posture: -2.8282 Episode_Reward/standstill: -6.7329 Episode_Reward/action_rate_l2: -0.5773 Episode_Reward/joint_torques_l2: -0.1510 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0368 Episode_Reward/unnecessary_wheel_lift: -0.1521 Episode_Reward/rolling_leg_motion: -0.7207 Episode_Reward/obstacle_progress: +2.2847 Episode_Reward/wheel_body_intrusion: -0.3470 Episode_Reward/pair_body_intrusion: -0.3921 Curriculum/terrain_level_mean: +2.3044 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1027 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.5476 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.3690 Episode_Termination/illegal_nonwheel_contact: 0.0536 ============================================================================== ============================================================================== Learning iteration 12700/50000 Total steps: 52,019,200 Steps per second: 10,616 Iteration time: 0.386s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0847 Mean episode length: 1392.7 Actor loss: -9.3792 Critic loss: +0.0044 World model loss: +0.8414 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1031 Episode_Reward/track_angular_velocity: +52.5155 Episode_Reward/upright: +27.0856 Episode_Reward/leg_posture: -2.8978 Episode_Reward/standstill: -6.8301 Episode_Reward/action_rate_l2: -0.6130 Episode_Reward/joint_torques_l2: -0.1532 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.1455 Episode_Reward/rolling_leg_motion: -0.7555 Episode_Reward/obstacle_progress: +2.4153 Episode_Reward/wheel_body_intrusion: -0.3377 Episode_Reward/pair_body_intrusion: -0.3652 Curriculum/terrain_level_mean: +2.3096 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1035 Metrics/base_velocity/error_vel_yaw: 0.1385 Episode_Termination/time_out: 0.5750 Episode_Termination/fell_over: 0.0187 Episode_Termination/course_exit: 0.3750 Episode_Termination/illegal_nonwheel_contact: 0.0437 ============================================================================== ============================================================================== Learning iteration 12750/50000 Total steps: 52,224,000 Steps per second: 10,630 Iteration time: 0.385s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1395.0 Actor loss: -9.3616 Critic loss: +0.0055 World model loss: +0.8966 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3591 Episode_Reward/track_angular_velocity: +52.7006 Episode_Reward/upright: +27.1624 Episode_Reward/leg_posture: -2.8770 Episode_Reward/standstill: -6.9255 Episode_Reward/action_rate_l2: -0.5981 Episode_Reward/joint_torques_l2: -0.1541 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0272 Episode_Reward/unnecessary_wheel_lift: -0.1283 Episode_Reward/rolling_leg_motion: -0.7444 Episode_Reward/obstacle_progress: +2.4708 Episode_Reward/wheel_body_intrusion: -0.3340 Episode_Reward/pair_body_intrusion: -0.3932 Curriculum/terrain_level_mean: +2.3123 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1023 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.5786 Episode_Termination/fell_over: 0.0252 Episode_Termination/course_exit: 0.3585 Episode_Termination/illegal_nonwheel_contact: 0.0503 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1043] ============================================================================== Learning iteration 12800/50000 Total steps: 52,428,800 Steps per second: 10,643 Iteration time: 0.385s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1381.6 Actor loss: -9.4209 Critic loss: +0.0072 World model loss: +0.8494 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9339 Episode_Reward/track_angular_velocity: +52.1282 Episode_Reward/upright: +26.8624 Episode_Reward/leg_posture: -2.8803 Episode_Reward/standstill: -6.7596 Episode_Reward/action_rate_l2: -0.5823 Episode_Reward/joint_torques_l2: -0.1537 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.1105 Episode_Reward/rolling_leg_motion: -0.7187 Episode_Reward/obstacle_progress: +2.4638 Episode_Reward/wheel_body_intrusion: -0.3614 Episode_Reward/pair_body_intrusion: -0.3937 Curriculum/terrain_level_mean: +2.3188 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1018 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.5694 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.3958 Episode_Termination/illegal_nonwheel_contact: 0.0139 ============================================================================== ============================================================================== Learning iteration 12850/50000 Total steps: 52,633,600 Steps per second: 10,656 Iteration time: 0.384s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1389.8 Actor loss: -9.3756 Critic loss: +0.0043 World model loss: +0.8819 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2085 Episode_Reward/track_angular_velocity: +52.4835 Episode_Reward/upright: +27.0155 Episode_Reward/leg_posture: -2.9332 Episode_Reward/standstill: -6.7930 Episode_Reward/action_rate_l2: -0.5842 Episode_Reward/joint_torques_l2: -0.1551 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.1205 Episode_Reward/rolling_leg_motion: -0.7485 Episode_Reward/obstacle_progress: +2.4990 Episode_Reward/wheel_body_intrusion: -0.3528 Episode_Reward/pair_body_intrusion: -0.4067 Curriculum/terrain_level_mean: +2.3247 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1021 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.5211 Episode_Termination/fell_over: 0.0352 Episode_Termination/course_exit: 0.4225 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== ============================================================================== Learning iteration 12900/50000 Total steps: 52,838,400 Steps per second: 10,670 Iteration time: 0.384s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0851 Mean episode length: 1442.0 Actor loss: -9.4251 Critic loss: +0.0042 World model loss: +0.8984 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0684 Episode_Reward/track_angular_velocity: +54.4105 Episode_Reward/upright: +28.0131 Episode_Reward/leg_posture: -3.0709 Episode_Reward/standstill: -6.7676 Episode_Reward/action_rate_l2: -0.5963 Episode_Reward/joint_torques_l2: -0.1617 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0141 Episode_Reward/unnecessary_wheel_lift: -0.1197 Episode_Reward/rolling_leg_motion: -0.7761 Episode_Reward/obstacle_progress: +2.5834 Episode_Reward/wheel_body_intrusion: -0.3820 Episode_Reward/pair_body_intrusion: -0.4318 Curriculum/terrain_level_mean: +2.3286 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1016 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.6284 Episode_Termination/fell_over: 0.0135 Episode_Termination/course_exit: 0.3378 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1674] ============================================================================== Learning iteration 12950/50000 Total steps: 53,043,200 Steps per second: 10,683 Iteration time: 0.383s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0850 Mean episode length: 1312.9 Actor loss: -9.4029 Critic loss: +0.0054 World model loss: +0.8799 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7696 Episode_Reward/track_angular_velocity: +49.5532 Episode_Reward/upright: +25.4848 Episode_Reward/leg_posture: -2.7728 Episode_Reward/standstill: -6.5933 Episode_Reward/action_rate_l2: -0.5569 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0256 Episode_Reward/unnecessary_wheel_lift: -0.1102 Episode_Reward/rolling_leg_motion: -0.7180 Episode_Reward/obstacle_progress: +2.3596 Episode_Reward/wheel_body_intrusion: -0.3269 Episode_Reward/pair_body_intrusion: -0.3659 Curriculum/terrain_level_mean: +2.3330 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1006 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.4150 Episode_Termination/fell_over: 0.0408 Episode_Termination/course_exit: 0.5170 Episode_Termination/illegal_nonwheel_contact: 0.0408 ============================================================================== ============================================================================== Learning iteration 13000/50000 Total steps: 53,248,000 Steps per second: 10,695 Iteration time: 0.383s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0851 Mean episode length: 1316.4 Actor loss: -9.4073 Critic loss: +0.0050 World model loss: +0.9043 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0104 Episode_Reward/track_angular_velocity: +49.6411 Episode_Reward/upright: +25.5450 Episode_Reward/leg_posture: -2.8106 Episode_Reward/standstill: -6.4176 Episode_Reward/action_rate_l2: -0.5609 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.1191 Episode_Reward/rolling_leg_motion: -0.7293 Episode_Reward/obstacle_progress: +2.3633 Episode_Reward/wheel_body_intrusion: -0.3410 Episode_Reward/pair_body_intrusion: -0.4479 Curriculum/terrain_level_mean: +2.3386 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0997 Metrics/base_velocity/error_vel_yaw: 0.1389 Episode_Termination/time_out: 0.4859 Episode_Termination/fell_over: 0.0352 Episode_Termination/course_exit: 0.4437 Episode_Termination/illegal_nonwheel_contact: 0.0563 ============================================================================== ============================================================================== Learning iteration 13050/50000 Total steps: 53,452,800 Steps per second: 10,708 Iteration time: 0.383s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0853 Mean episode length: 1338.8 Actor loss: -9.4215 Critic loss: +0.0054 World model loss: +0.9093 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7540 Episode_Reward/track_angular_velocity: +50.4710 Episode_Reward/upright: +25.9865 Episode_Reward/leg_posture: -2.8427 Episode_Reward/standstill: -6.3398 Episode_Reward/action_rate_l2: -0.5833 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1195 Episode_Reward/rolling_leg_motion: -0.7372 Episode_Reward/obstacle_progress: +2.4878 Episode_Reward/wheel_body_intrusion: -0.3385 Episode_Reward/pair_body_intrusion: -0.4140 Curriculum/terrain_level_mean: +2.3430 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1004 Metrics/base_velocity/error_vel_yaw: 0.1392 Episode_Termination/time_out: 0.5159 Episode_Termination/fell_over: 0.0079 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== ============================================================================== Learning iteration 13100/50000 Total steps: 53,657,600 Steps per second: 10,722 Iteration time: 0.382s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0849 Mean episode length: 1392.7 Actor loss: -9.4271 Critic loss: +0.0067 World model loss: +0.9341 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3301 Episode_Reward/track_angular_velocity: +52.4424 Episode_Reward/upright: +26.9879 Episode_Reward/leg_posture: -2.9315 Episode_Reward/standstill: -6.7650 Episode_Reward/action_rate_l2: -0.6188 Episode_Reward/joint_torques_l2: -0.1593 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0190 Episode_Reward/unnecessary_wheel_lift: -0.1302 Episode_Reward/rolling_leg_motion: -0.7809 Episode_Reward/obstacle_progress: +2.6224 Episode_Reward/wheel_body_intrusion: -0.3287 Episode_Reward/pair_body_intrusion: -0.4445 Curriculum/terrain_level_mean: +2.3467 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1033 Metrics/base_velocity/error_vel_yaw: 0.1409 Episode_Termination/time_out: 0.5349 Episode_Termination/fell_over: 0.0388 Episode_Termination/course_exit: 0.4109 Episode_Termination/illegal_nonwheel_contact: 0.0310 ============================================================================== ============================================================================== Learning iteration 13150/50000 Total steps: 53,862,400 Steps per second: 10,735 Iteration time: 0.382s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0851 Mean episode length: 1324.8 Actor loss: -9.3938 Critic loss: +0.0093 World model loss: +0.9053 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1632 Episode_Reward/track_angular_velocity: +49.8904 Episode_Reward/upright: +25.6653 Episode_Reward/leg_posture: -2.7830 Episode_Reward/standstill: -6.4881 Episode_Reward/action_rate_l2: -0.5913 Episode_Reward/joint_torques_l2: -0.1514 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.1058 Episode_Reward/rolling_leg_motion: -0.7426 Episode_Reward/obstacle_progress: +2.5270 Episode_Reward/wheel_body_intrusion: -0.3178 Episode_Reward/pair_body_intrusion: -0.3431 Curriculum/terrain_level_mean: +2.3545 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1013 Metrics/base_velocity/error_vel_yaw: 0.1406 Episode_Termination/time_out: 0.4604 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.4892 Episode_Termination/illegal_nonwheel_contact: 0.0216 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[774] RECOVERY reset 1 invalid environments; first IDs=[682] ============================================================================== Learning iteration 13200/50000 Total steps: 54,067,200 Steps per second: 10,747 Iteration time: 0.381s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0852 Mean episode length: 1388.0 Actor loss: -9.4403 Critic loss: +0.0041 World model loss: +0.9226 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4382 Episode_Reward/track_angular_velocity: +52.2415 Episode_Reward/upright: +26.8729 Episode_Reward/leg_posture: -2.9395 Episode_Reward/standstill: -6.6545 Episode_Reward/action_rate_l2: -0.6200 Episode_Reward/joint_torques_l2: -0.1597 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.1276 Episode_Reward/rolling_leg_motion: -0.8002 Episode_Reward/obstacle_progress: +2.7167 Episode_Reward/wheel_body_intrusion: -0.3378 Episode_Reward/pair_body_intrusion: -0.3649 Curriculum/terrain_level_mean: +2.3586 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1009 Metrics/base_velocity/error_vel_yaw: 0.1410 Episode_Termination/time_out: 0.5347 Episode_Termination/fell_over: 0.0417 Episode_Termination/course_exit: 0.3958 Episode_Termination/illegal_nonwheel_contact: 0.0486 ============================================================================== ============================================================================== Learning iteration 13250/50000 Total steps: 54,272,000 Steps per second: 10,760 Iteration time: 0.381s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0854 Mean episode length: 1409.1 Actor loss: -9.4008 Critic loss: +0.0062 World model loss: +0.9198 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3157 Episode_Reward/track_angular_velocity: +53.1146 Episode_Reward/upright: +27.3131 Episode_Reward/leg_posture: -2.9935 Episode_Reward/standstill: -6.7054 Episode_Reward/action_rate_l2: -0.6379 Episode_Reward/joint_torques_l2: -0.1609 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.1248 Episode_Reward/rolling_leg_motion: -0.7989 Episode_Reward/obstacle_progress: +2.8012 Episode_Reward/wheel_body_intrusion: -0.3555 Episode_Reward/pair_body_intrusion: -0.3803 Curriculum/terrain_level_mean: +2.3650 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0997 Metrics/base_velocity/error_vel_yaw: 0.1397 Episode_Termination/time_out: 0.5724 Episode_Termination/fell_over: 0.0345 Episode_Termination/course_exit: 0.3862 Episode_Termination/illegal_nonwheel_contact: 0.0276 ============================================================================== ============================================================================== Learning iteration 13300/50000 Total steps: 54,476,800 Steps per second: 10,773 Iteration time: 0.380s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0854 Mean episode length: 1320.7 Actor loss: -9.4502 Critic loss: +0.0069 World model loss: +0.9435 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2364 Episode_Reward/track_angular_velocity: +49.8066 Episode_Reward/upright: +25.5815 Episode_Reward/leg_posture: -2.7971 Episode_Reward/standstill: -6.3839 Episode_Reward/action_rate_l2: -0.5863 Episode_Reward/joint_torques_l2: -0.1511 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.1072 Episode_Reward/rolling_leg_motion: -0.7495 Episode_Reward/obstacle_progress: +2.6648 Episode_Reward/wheel_body_intrusion: -0.3319 Episode_Reward/pair_body_intrusion: -0.3858 Curriculum/terrain_level_mean: +2.3735 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0995 Metrics/base_velocity/error_vel_yaw: 0.1391 Episode_Termination/time_out: 0.4527 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== ============================================================================== Learning iteration 13350/50000 Total steps: 54,681,600 Steps per second: 10,786 Iteration time: 0.380s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0848 Mean episode length: 1375.0 Actor loss: -9.4448 Critic loss: +0.0032 World model loss: +0.9433 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0793 Episode_Reward/track_angular_velocity: +51.7750 Episode_Reward/upright: +26.6247 Episode_Reward/leg_posture: -2.9497 Episode_Reward/standstill: -7.0276 Episode_Reward/action_rate_l2: -0.6154 Episode_Reward/joint_torques_l2: -0.1573 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0430 Episode_Reward/unnecessary_wheel_lift: -0.1168 Episode_Reward/rolling_leg_motion: -0.7768 Episode_Reward/obstacle_progress: +2.6439 Episode_Reward/wheel_body_intrusion: -0.3418 Episode_Reward/pair_body_intrusion: -0.4331 Curriculum/terrain_level_mean: +2.3799 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.1003 Metrics/base_velocity/error_vel_yaw: 0.1408 Episode_Termination/time_out: 0.4730 Episode_Termination/fell_over: 0.0541 Episode_Termination/course_exit: 0.4189 Episode_Termination/illegal_nonwheel_contact: 0.0743 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2258] ============================================================================== Learning iteration 13400/50000 Total steps: 54,886,400 Steps per second: 10,798 Iteration time: 0.379s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0854 Mean episode length: 1397.9 Actor loss: -9.4312 Critic loss: +0.0048 World model loss: +0.9224 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9911 Episode_Reward/track_angular_velocity: +52.6757 Episode_Reward/upright: +27.0801 Episode_Reward/leg_posture: -2.9313 Episode_Reward/standstill: -6.7542 Episode_Reward/action_rate_l2: -0.6239 Episode_Reward/joint_torques_l2: -0.1598 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0246 Episode_Reward/unnecessary_wheel_lift: -0.1330 Episode_Reward/rolling_leg_motion: -0.7961 Episode_Reward/obstacle_progress: +2.7086 Episode_Reward/wheel_body_intrusion: -0.3079 Episode_Reward/pair_body_intrusion: -0.3464 Curriculum/terrain_level_mean: +2.3865 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0993 Metrics/base_velocity/error_vel_yaw: 0.1399 Episode_Termination/time_out: 0.5448 Episode_Termination/fell_over: 0.0483 Episode_Termination/course_exit: 0.3931 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== ============================================================================== Learning iteration 13450/50000 Total steps: 55,091,200 Steps per second: 10,811 Iteration time: 0.379s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0851 Mean episode length: 1358.5 Actor loss: -9.4611 Critic loss: +0.0033 World model loss: +0.9430 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5811 Episode_Reward/track_angular_velocity: +51.1605 Episode_Reward/upright: +26.3276 Episode_Reward/leg_posture: -2.8514 Episode_Reward/standstill: -6.8684 Episode_Reward/action_rate_l2: -0.6173 Episode_Reward/joint_torques_l2: -0.1550 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0318 Episode_Reward/unnecessary_wheel_lift: -0.1251 Episode_Reward/rolling_leg_motion: -0.7640 Episode_Reward/obstacle_progress: +2.6576 Episode_Reward/wheel_body_intrusion: -0.3178 Episode_Reward/pair_body_intrusion: -0.3498 Curriculum/terrain_level_mean: +2.3921 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0996 Metrics/base_velocity/error_vel_yaw: 0.1408 Episode_Termination/time_out: 0.5143 Episode_Termination/fell_over: 0.0571 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0643 ============================================================================== ============================================================================== Learning iteration 13500/50000 Total steps: 55,296,000 Steps per second: 10,824 Iteration time: 0.378s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0856 Mean episode length: 1355.1 Actor loss: -9.4415 Critic loss: +0.0043 World model loss: +0.9627 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7348 Episode_Reward/track_angular_velocity: +51.1284 Episode_Reward/upright: +26.3012 Episode_Reward/leg_posture: -2.8334 Episode_Reward/standstill: -6.6187 Episode_Reward/action_rate_l2: -0.6018 Episode_Reward/joint_torques_l2: -0.1540 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.1122 Episode_Reward/rolling_leg_motion: -0.7705 Episode_Reward/obstacle_progress: +2.7048 Episode_Reward/wheel_body_intrusion: -0.3056 Episode_Reward/pair_body_intrusion: -0.3931 Curriculum/terrain_level_mean: +2.3979 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1383 Episode_Termination/time_out: 0.5461 Episode_Termination/fell_over: 0.0329 Episode_Termination/course_exit: 0.3947 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_013500.mp4 ============================================================================== Learning iteration 13550/50000 Total steps: 55,500,800 Steps per second: 10,823 Iteration time: 0.378s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0855 Mean episode length: 1435.9 Actor loss: -9.4487 Critic loss: +0.0042 World model loss: +0.9218 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.5912 Episode_Reward/track_angular_velocity: +54.1197 Episode_Reward/upright: +27.8582 Episode_Reward/leg_posture: -3.0575 Episode_Reward/standstill: -6.9642 Episode_Reward/action_rate_l2: -0.6306 Episode_Reward/joint_torques_l2: -0.1638 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0168 Episode_Reward/unnecessary_wheel_lift: -0.1272 Episode_Reward/rolling_leg_motion: -0.8379 Episode_Reward/obstacle_progress: +2.7997 Episode_Reward/wheel_body_intrusion: -0.3574 Episode_Reward/pair_body_intrusion: -0.4613 Curriculum/terrain_level_mean: +2.4041 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0970 Metrics/base_velocity/error_vel_yaw: 0.1397 Episode_Termination/time_out: 0.5779 Episode_Termination/fell_over: 0.0195 Episode_Termination/course_exit: 0.3831 Episode_Termination/illegal_nonwheel_contact: 0.0260 ============================================================================== ============================================================================== Learning iteration 13600/50000 Total steps: 55,705,600 Steps per second: 10,836 Iteration time: 0.378s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1364.8 Actor loss: -9.4574 Critic loss: +0.0066 World model loss: +0.9413 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0464 Episode_Reward/track_angular_velocity: +51.4907 Episode_Reward/upright: +26.4853 Episode_Reward/leg_posture: -2.8561 Episode_Reward/standstill: -6.7066 Episode_Reward/action_rate_l2: -0.6054 Episode_Reward/joint_torques_l2: -0.1548 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.1304 Episode_Reward/rolling_leg_motion: -0.7804 Episode_Reward/obstacle_progress: +2.6677 Episode_Reward/wheel_body_intrusion: -0.3147 Episode_Reward/pair_body_intrusion: -0.3433 Curriculum/terrain_level_mean: +2.4116 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1385 Episode_Termination/time_out: 0.4797 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.4865 Episode_Termination/illegal_nonwheel_contact: 0.0203 ============================================================================== ============================================================================== Learning iteration 13650/50000 Total steps: 55,910,400 Steps per second: 10,848 Iteration time: 0.378s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0854 Mean episode length: 1370.8 Actor loss: -9.4525 Critic loss: +0.0044 World model loss: +0.9045 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3014 Episode_Reward/track_angular_velocity: +51.6511 Episode_Reward/upright: +26.5783 Episode_Reward/leg_posture: -2.8818 Episode_Reward/standstill: -6.8886 Episode_Reward/action_rate_l2: -0.6184 Episode_Reward/joint_torques_l2: -0.1560 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0295 Episode_Reward/unnecessary_wheel_lift: -0.1354 Episode_Reward/rolling_leg_motion: -0.8073 Episode_Reward/obstacle_progress: +2.7166 Episode_Reward/wheel_body_intrusion: -0.2937 Episode_Reward/pair_body_intrusion: -0.3626 Curriculum/terrain_level_mean: +2.4177 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1398 Episode_Termination/time_out: 0.5094 Episode_Termination/fell_over: 0.0503 Episode_Termination/course_exit: 0.4214 Episode_Termination/illegal_nonwheel_contact: 0.0377 ============================================================================== ============================================================================== Learning iteration 13700/50000 Total steps: 56,115,200 Steps per second: 10,860 Iteration time: 0.377s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1368.2 Actor loss: -9.4239 Critic loss: +0.0039 World model loss: +0.9292 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1751 Episode_Reward/track_angular_velocity: +51.6314 Episode_Reward/upright: +26.5154 Episode_Reward/leg_posture: -2.8827 Episode_Reward/standstill: -6.8002 Episode_Reward/action_rate_l2: -0.5941 Episode_Reward/joint_torques_l2: -0.1576 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1239 Episode_Reward/rolling_leg_motion: -0.7905 Episode_Reward/obstacle_progress: +2.7764 Episode_Reward/wheel_body_intrusion: -0.3001 Episode_Reward/pair_body_intrusion: -0.3665 Curriculum/terrain_level_mean: +2.4214 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1378 Episode_Termination/time_out: 0.4891 Episode_Termination/fell_over: 0.0365 Episode_Termination/course_exit: 0.4526 Episode_Termination/illegal_nonwheel_contact: 0.0219 ============================================================================== ============================================================================== Learning iteration 13750/50000 Total steps: 56,320,000 Steps per second: 10,872 Iteration time: 0.377s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0855 Mean episode length: 1328.1 Actor loss: -9.4663 Critic loss: +0.0032 World model loss: +0.9019 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8106 Episode_Reward/track_angular_velocity: +50.1052 Episode_Reward/upright: +25.7241 Episode_Reward/leg_posture: -2.8112 Episode_Reward/standstill: -6.6777 Episode_Reward/action_rate_l2: -0.5809 Episode_Reward/joint_torques_l2: -0.1527 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.1404 Episode_Reward/rolling_leg_motion: -0.7754 Episode_Reward/obstacle_progress: +2.6604 Episode_Reward/wheel_body_intrusion: -0.2860 Episode_Reward/pair_body_intrusion: -0.3053 Curriculum/terrain_level_mean: +2.4290 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0970 Metrics/base_velocity/error_vel_yaw: 0.1380 Episode_Termination/time_out: 0.4803 Episode_Termination/fell_over: 0.0329 Episode_Termination/course_exit: 0.4803 Episode_Termination/illegal_nonwheel_contact: 0.0197 ============================================================================== ============================================================================== Learning iteration 13800/50000 Total steps: 56,524,800 Steps per second: 10,884 Iteration time: 0.376s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0855 Mean episode length: 1381.1 Actor loss: -9.4793 Critic loss: +0.0045 World model loss: +0.8919 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7053 Episode_Reward/track_angular_velocity: +52.1273 Episode_Reward/upright: +26.7370 Episode_Reward/leg_posture: -2.9498 Episode_Reward/standstill: -7.0120 Episode_Reward/action_rate_l2: -0.6100 Episode_Reward/joint_torques_l2: -0.1578 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0378 Episode_Reward/unnecessary_wheel_lift: -0.1496 Episode_Reward/rolling_leg_motion: -0.7985 Episode_Reward/obstacle_progress: +2.7790 Episode_Reward/wheel_body_intrusion: -0.3058 Episode_Reward/pair_body_intrusion: -0.3079 Curriculum/terrain_level_mean: +2.4336 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0353 Episode_Termination/course_exit: 0.4294 Episode_Termination/illegal_nonwheel_contact: 0.0647 ============================================================================== ============================================================================== Learning iteration 13850/50000 Total steps: 56,729,600 Steps per second: 10,896 Iteration time: 0.376s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1320.9 Actor loss: -9.4916 Critic loss: +0.0042 World model loss: +0.9088 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6314 Episode_Reward/track_angular_velocity: +49.9059 Episode_Reward/upright: +25.5677 Episode_Reward/leg_posture: -2.7526 Episode_Reward/standstill: -6.5781 Episode_Reward/action_rate_l2: -0.5995 Episode_Reward/joint_torques_l2: -0.1509 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0335 Episode_Reward/unnecessary_wheel_lift: -0.1402 Episode_Reward/rolling_leg_motion: -0.7893 Episode_Reward/obstacle_progress: +2.6446 Episode_Reward/wheel_body_intrusion: -0.2837 Episode_Reward/pair_body_intrusion: -0.3131 Curriculum/terrain_level_mean: +2.4412 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0964 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4606 Episode_Termination/fell_over: 0.0485 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0545 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1948] ============================================================================== Learning iteration 13900/50000 Total steps: 56,934,400 Steps per second: 10,908 Iteration time: 0.375s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1405.7 Actor loss: -9.4761 Critic loss: +0.0044 World model loss: +0.8977 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8224 Episode_Reward/track_angular_velocity: +53.1254 Episode_Reward/upright: +27.2290 Episode_Reward/leg_posture: -2.9385 Episode_Reward/standstill: -6.6508 Episode_Reward/action_rate_l2: -0.6254 Episode_Reward/joint_torques_l2: -0.1606 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0192 Episode_Reward/unnecessary_wheel_lift: -0.1252 Episode_Reward/rolling_leg_motion: -0.8015 Episode_Reward/obstacle_progress: +2.8634 Episode_Reward/wheel_body_intrusion: -0.3021 Episode_Reward/pair_body_intrusion: -0.3565 Curriculum/terrain_level_mean: +2.4441 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.5147 Episode_Termination/fell_over: 0.0368 Episode_Termination/course_exit: 0.4338 Episode_Termination/illegal_nonwheel_contact: 0.0221 ============================================================================== ============================================================================== Learning iteration 13950/50000 Total steps: 57,139,200 Steps per second: 10,920 Iteration time: 0.375s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0856 Mean episode length: 1385.5 Actor loss: -9.4493 Critic loss: +0.0034 World model loss: +0.8534 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9625 Episode_Reward/track_angular_velocity: +52.2509 Episode_Reward/upright: +26.8508 Episode_Reward/leg_posture: -2.8816 Episode_Reward/standstill: -6.9913 Episode_Reward/action_rate_l2: -0.6003 Episode_Reward/joint_torques_l2: -0.1573 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0325 Episode_Reward/unnecessary_wheel_lift: -0.1483 Episode_Reward/rolling_leg_motion: -0.7995 Episode_Reward/obstacle_progress: +2.7767 Episode_Reward/wheel_body_intrusion: -0.2952 Episode_Reward/pair_body_intrusion: -0.3078 Curriculum/terrain_level_mean: +2.4482 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0955 Metrics/base_velocity/error_vel_yaw: 0.1388 Episode_Termination/time_out: 0.5093 Episode_Termination/fell_over: 0.0435 Episode_Termination/course_exit: 0.4161 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== ============================================================================== Learning iteration 14000/50000 Total steps: 57,344,000 Steps per second: 10,932 Iteration time: 0.375s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1420.5 Actor loss: -9.4792 Critic loss: +0.0073 World model loss: +0.9023 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.4559 Episode_Reward/track_angular_velocity: +53.6509 Episode_Reward/upright: +27.5495 Episode_Reward/leg_posture: -2.9551 Episode_Reward/standstill: -6.8720 Episode_Reward/action_rate_l2: -0.6152 Episode_Reward/joint_torques_l2: -0.1605 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0194 Episode_Reward/unnecessary_wheel_lift: -0.1308 Episode_Reward/rolling_leg_motion: -0.8236 Episode_Reward/obstacle_progress: +2.8600 Episode_Reward/wheel_body_intrusion: -0.3141 Episode_Reward/pair_body_intrusion: -0.3677 Curriculum/terrain_level_mean: +2.4536 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.5419 Episode_Termination/fell_over: 0.0258 Episode_Termination/course_exit: 0.4129 Episode_Termination/illegal_nonwheel_contact: 0.0387 ============================================================================== ============================================================================== Learning iteration 14050/50000 Total steps: 57,548,800 Steps per second: 10,944 Iteration time: 0.374s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0861 Mean episode length: 1349.5 Actor loss: -9.4592 Critic loss: +0.0108 World model loss: +0.9051 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9480 Episode_Reward/track_angular_velocity: +50.9779 Episode_Reward/upright: +26.2241 Episode_Reward/leg_posture: -2.7973 Episode_Reward/standstill: -6.4555 Episode_Reward/action_rate_l2: -0.5831 Episode_Reward/joint_torques_l2: -0.1497 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.1157 Episode_Reward/rolling_leg_motion: -0.7755 Episode_Reward/obstacle_progress: +2.6629 Episode_Reward/wheel_body_intrusion: -0.3087 Episode_Reward/pair_body_intrusion: -0.3593 Curriculum/terrain_level_mean: +2.4602 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4408 Episode_Termination/fell_over: 0.0263 Episode_Termination/course_exit: 0.5132 Episode_Termination/illegal_nonwheel_contact: 0.0197 ============================================================================== ============================================================================== Learning iteration 14100/50000 Total steps: 57,753,600 Steps per second: 10,956 Iteration time: 0.374s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0858 Mean episode length: 1330.9 Actor loss: -9.4661 Critic loss: +0.0057 World model loss: +0.8784 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1961 Episode_Reward/track_angular_velocity: +50.2810 Episode_Reward/upright: +25.8318 Episode_Reward/leg_posture: -2.7966 Episode_Reward/standstill: -6.5560 Episode_Reward/action_rate_l2: -0.5850 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0286 Episode_Reward/unnecessary_wheel_lift: -0.1342 Episode_Reward/rolling_leg_motion: -0.7782 Episode_Reward/obstacle_progress: +2.5777 Episode_Reward/wheel_body_intrusion: -0.3103 Episode_Reward/pair_body_intrusion: -0.3238 Curriculum/terrain_level_mean: +2.4656 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4899 Episode_Termination/fell_over: 0.0470 Episode_Termination/course_exit: 0.4362 Episode_Termination/illegal_nonwheel_contact: 0.0470 ============================================================================== ============================================================================== Learning iteration 14150/50000 Total steps: 57,958,400 Steps per second: 10,968 Iteration time: 0.373s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0860 Mean episode length: 1415.9 Actor loss: -9.4976 Critic loss: +0.0038 World model loss: +0.8249 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2073 Episode_Reward/track_angular_velocity: +53.5413 Episode_Reward/upright: +27.5126 Episode_Reward/leg_posture: -2.9438 Episode_Reward/standstill: -6.8862 Episode_Reward/action_rate_l2: -0.6195 Episode_Reward/joint_torques_l2: -0.1567 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1421 Episode_Reward/rolling_leg_motion: -0.8322 Episode_Reward/obstacle_progress: +2.7037 Episode_Reward/wheel_body_intrusion: -0.3106 Episode_Reward/pair_body_intrusion: -0.3355 Curriculum/terrain_level_mean: +2.4724 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5061 Episode_Termination/fell_over: 0.0244 Episode_Termination/course_exit: 0.4451 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1582] ============================================================================== Learning iteration 14200/50000 Total steps: 58,163,200 Steps per second: 10,979 Iteration time: 0.373s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0856 Mean episode length: 1370.2 Actor loss: -9.4899 Critic loss: +0.0066 World model loss: +0.8660 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5742 Episode_Reward/track_angular_velocity: +51.8021 Episode_Reward/upright: +26.5785 Episode_Reward/leg_posture: -2.8773 Episode_Reward/standstill: -7.0216 Episode_Reward/action_rate_l2: -0.5881 Episode_Reward/joint_torques_l2: -0.1533 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.1446 Episode_Reward/rolling_leg_motion: -0.7970 Episode_Reward/obstacle_progress: +2.6256 Episode_Reward/wheel_body_intrusion: -0.3039 Episode_Reward/pair_body_intrusion: -0.3283 Curriculum/terrain_level_mean: +2.4783 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4872 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4615 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== ============================================================================== Learning iteration 14250/50000 Total steps: 58,368,000 Steps per second: 10,991 Iteration time: 0.373s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0856 Mean episode length: 1385.0 Actor loss: -9.4738 Critic loss: +0.0039 World model loss: +0.8444 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0806 Episode_Reward/track_angular_velocity: +52.3894 Episode_Reward/upright: +26.8913 Episode_Reward/leg_posture: -2.8703 Episode_Reward/standstill: -7.2912 Episode_Reward/action_rate_l2: -0.6137 Episode_Reward/joint_torques_l2: -0.1553 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0243 Episode_Reward/unnecessary_wheel_lift: -0.1326 Episode_Reward/rolling_leg_motion: -0.8140 Episode_Reward/obstacle_progress: +2.6302 Episode_Reward/wheel_body_intrusion: -0.2861 Episode_Reward/pair_body_intrusion: -0.3160 Curriculum/terrain_level_mean: +2.4856 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5159 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4268 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== ============================================================================== Learning iteration 14300/50000 Total steps: 58,572,800 Steps per second: 11,002 Iteration time: 0.372s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0858 Mean episode length: 1416.6 Actor loss: -9.4994 Critic loss: +0.0057 World model loss: +0.8593 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0685 Episode_Reward/track_angular_velocity: +53.5856 Episode_Reward/upright: +27.5165 Episode_Reward/leg_posture: -2.9782 Episode_Reward/standstill: -6.9743 Episode_Reward/action_rate_l2: -0.6222 Episode_Reward/joint_torques_l2: -0.1596 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0194 Episode_Reward/unnecessary_wheel_lift: -0.1393 Episode_Reward/rolling_leg_motion: -0.8373 Episode_Reward/obstacle_progress: +2.7261 Episode_Reward/wheel_body_intrusion: -0.3146 Episode_Reward/pair_body_intrusion: -0.3007 Curriculum/terrain_level_mean: +2.4932 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0955 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5376 Episode_Termination/fell_over: 0.0289 Episode_Termination/course_exit: 0.4220 Episode_Termination/illegal_nonwheel_contact: 0.0173 ============================================================================== ============================================================================== Learning iteration 14350/50000 Total steps: 58,777,600 Steps per second: 11,013 Iteration time: 0.372s Collection time: 0.075s Learning time: 0.193s Mean reward: +0.0860 Mean episode length: 1378.4 Actor loss: -9.4914 Critic loss: +0.0052 World model loss: +0.7984 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8628 Episode_Reward/track_angular_velocity: +52.0540 Episode_Reward/upright: +26.7711 Episode_Reward/leg_posture: -2.8396 Episode_Reward/standstill: -6.7777 Episode_Reward/action_rate_l2: -0.6178 Episode_Reward/joint_torques_l2: -0.1529 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.1198 Episode_Reward/rolling_leg_motion: -0.7896 Episode_Reward/obstacle_progress: +2.7096 Episode_Reward/wheel_body_intrusion: -0.2988 Episode_Reward/pair_body_intrusion: -0.3053 Curriculum/terrain_level_mean: +2.4990 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5369 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.4161 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 14400/50000 Total steps: 58,982,400 Steps per second: 11,025 Iteration time: 0.372s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1299.8 Actor loss: -9.4924 Critic loss: +0.0028 World model loss: +0.8269 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0219 Episode_Reward/track_angular_velocity: +49.0809 Episode_Reward/upright: +25.2355 Episode_Reward/leg_posture: -2.6686 Episode_Reward/standstill: -6.0518 Episode_Reward/action_rate_l2: -0.5988 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0228 Episode_Reward/unnecessary_wheel_lift: -0.1320 Episode_Reward/rolling_leg_motion: -0.7504 Episode_Reward/obstacle_progress: +2.6224 Episode_Reward/wheel_body_intrusion: -0.2716 Episode_Reward/pair_body_intrusion: -0.2827 Curriculum/terrain_level_mean: +2.5090 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1375 Episode_Termination/time_out: 0.4596 Episode_Termination/fell_over: 0.0202 Episode_Termination/course_exit: 0.4949 Episode_Termination/illegal_nonwheel_contact: 0.0404 ============================================================================== ============================================================================== Learning iteration 14450/50000 Total steps: 59,187,200 Steps per second: 11,036 Iteration time: 0.371s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0863 Mean episode length: 1389.9 Actor loss: -9.4857 Critic loss: +0.0038 World model loss: +0.8417 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1998 Episode_Reward/track_angular_velocity: +52.4671 Episode_Reward/upright: +26.9972 Episode_Reward/leg_posture: -2.8272 Episode_Reward/standstill: -6.2739 Episode_Reward/action_rate_l2: -0.6268 Episode_Reward/joint_torques_l2: -0.1531 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0136 Episode_Reward/unnecessary_wheel_lift: -0.1462 Episode_Reward/rolling_leg_motion: -0.8135 Episode_Reward/obstacle_progress: +2.8310 Episode_Reward/wheel_body_intrusion: -0.2858 Episode_Reward/pair_body_intrusion: -0.3395 Curriculum/terrain_level_mean: +2.5183 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.4916 Episode_Termination/fell_over: 0.0112 Episode_Termination/course_exit: 0.4804 Episode_Termination/illegal_nonwheel_contact: 0.0279 ============================================================================== ============================================================================== Learning iteration 14500/50000 Total steps: 59,392,000 Steps per second: 11,047 Iteration time: 0.371s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0866 Mean episode length: 1353.7 Actor loss: -9.5064 Critic loss: +0.0031 World model loss: +0.8357 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0166 Episode_Reward/track_angular_velocity: +51.2171 Episode_Reward/upright: +26.2862 Episode_Reward/leg_posture: -2.7804 Episode_Reward/standstill: -6.1155 Episode_Reward/action_rate_l2: -0.5942 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0026 Episode_Reward/unnecessary_wheel_lift: -0.1109 Episode_Reward/rolling_leg_motion: -0.7506 Episode_Reward/obstacle_progress: +2.7810 Episode_Reward/wheel_body_intrusion: -0.2851 Episode_Reward/pair_body_intrusion: -0.3239 Curriculum/terrain_level_mean: +2.5276 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5211 Episode_Termination/fell_over: 0.0070 Episode_Termination/course_exit: 0.4718 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 14550/50000 Total steps: 59,596,800 Steps per second: 11,058 Iteration time: 0.370s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1306.1 Actor loss: -9.4687 Critic loss: +0.0047 World model loss: +0.8378 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0785 Episode_Reward/track_angular_velocity: +49.4083 Episode_Reward/upright: +25.3454 Episode_Reward/leg_posture: -2.6942 Episode_Reward/standstill: -5.8166 Episode_Reward/action_rate_l2: -0.5896 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0230 Episode_Reward/unnecessary_wheel_lift: -0.1332 Episode_Reward/rolling_leg_motion: -0.7513 Episode_Reward/obstacle_progress: +2.6632 Episode_Reward/wheel_body_intrusion: -0.2831 Episode_Reward/pair_body_intrusion: -0.3289 Curriculum/terrain_level_mean: +2.5361 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0309 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0494 ============================================================================== ============================================================================== Learning iteration 14600/50000 Total steps: 59,801,600 Steps per second: 11,070 Iteration time: 0.370s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1328.7 Actor loss: -9.4642 Critic loss: +0.0063 World model loss: +0.8470 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8083 Episode_Reward/track_angular_velocity: +50.2063 Episode_Reward/upright: +25.7722 Episode_Reward/leg_posture: -2.6609 Episode_Reward/standstill: -6.1230 Episode_Reward/action_rate_l2: -0.6168 Episode_Reward/joint_torques_l2: -0.1455 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.1200 Episode_Reward/rolling_leg_motion: -0.7551 Episode_Reward/obstacle_progress: +2.6786 Episode_Reward/wheel_body_intrusion: -0.2661 Episode_Reward/pair_body_intrusion: -0.2835 Curriculum/terrain_level_mean: +2.5469 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0966 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4356 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.5337 Episode_Termination/illegal_nonwheel_contact: 0.0184 ============================================================================== ============================================================================== Learning iteration 14650/50000 Total steps: 60,006,400 Steps per second: 11,081 Iteration time: 0.370s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1342.1 Actor loss: -9.4762 Critic loss: +0.0037 World model loss: +0.8297 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3649 Episode_Reward/track_angular_velocity: +50.7136 Episode_Reward/upright: +26.0385 Episode_Reward/leg_posture: -2.7282 Episode_Reward/standstill: -6.1713 Episode_Reward/action_rate_l2: -0.6043 Episode_Reward/joint_torques_l2: -0.1479 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0236 Episode_Reward/unnecessary_wheel_lift: -0.1233 Episode_Reward/rolling_leg_motion: -0.7641 Episode_Reward/obstacle_progress: +2.7592 Episode_Reward/wheel_body_intrusion: -0.2804 Episode_Reward/pair_body_intrusion: -0.3182 Curriculum/terrain_level_mean: +2.5525 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0966 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4702 Episode_Termination/fell_over: 0.0397 Episode_Termination/course_exit: 0.4702 Episode_Termination/illegal_nonwheel_contact: 0.0397 ============================================================================== ============================================================================== Learning iteration 14700/50000 Total steps: 60,211,200 Steps per second: 11,092 Iteration time: 0.369s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1319.8 Actor loss: -9.4568 Critic loss: +0.0077 World model loss: +0.8190 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5215 Episode_Reward/track_angular_velocity: +49.8917 Episode_Reward/upright: +25.6137 Episode_Reward/leg_posture: -2.7000 Episode_Reward/standstill: -5.9612 Episode_Reward/action_rate_l2: -0.5905 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.1147 Episode_Reward/rolling_leg_motion: -0.7480 Episode_Reward/obstacle_progress: +2.6468 Episode_Reward/wheel_body_intrusion: -0.2652 Episode_Reward/pair_body_intrusion: -0.3333 Curriculum/terrain_level_mean: +2.5564 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0965 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4648 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.4577 Episode_Termination/illegal_nonwheel_contact: 0.0704 ============================================================================== ============================================================================== Learning iteration 14750/50000 Total steps: 60,416,000 Steps per second: 11,103 Iteration time: 0.369s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0863 Mean episode length: 1440.8 Actor loss: -9.5000 Critic loss: +0.0029 World model loss: +0.8109 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7175 Episode_Reward/track_angular_velocity: +54.4682 Episode_Reward/upright: +27.9528 Episode_Reward/leg_posture: -2.9440 Episode_Reward/standstill: -6.4283 Episode_Reward/action_rate_l2: -0.6446 Episode_Reward/joint_torques_l2: -0.1577 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0197 Episode_Reward/unnecessary_wheel_lift: -0.1205 Episode_Reward/rolling_leg_motion: -0.7907 Episode_Reward/obstacle_progress: +2.8903 Episode_Reward/wheel_body_intrusion: -0.2852 Episode_Reward/pair_body_intrusion: -0.3476 Curriculum/terrain_level_mean: +2.5627 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0968 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.4521 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 14800/50000 Total steps: 60,620,800 Steps per second: 11,114 Iteration time: 0.369s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0861 Mean episode length: 1273.3 Actor loss: -9.4654 Critic loss: +0.0036 World model loss: +0.8368 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.7451 Episode_Reward/track_angular_velocity: +48.1125 Episode_Reward/upright: +24.6657 Episode_Reward/leg_posture: -2.5901 Episode_Reward/standstill: -5.7663 Episode_Reward/action_rate_l2: -0.5766 Episode_Reward/joint_torques_l2: -0.1385 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0149 Episode_Reward/unnecessary_wheel_lift: -0.0986 Episode_Reward/rolling_leg_motion: -0.7014 Episode_Reward/obstacle_progress: +2.6088 Episode_Reward/wheel_body_intrusion: -0.2511 Episode_Reward/pair_body_intrusion: -0.2978 Curriculum/terrain_level_mean: +2.5706 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0977 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.3629 Episode_Termination/fell_over: 0.0242 Episode_Termination/course_exit: 0.5887 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== ============================================================================== Learning iteration 14850/50000 Total steps: 60,825,600 Steps per second: 11,125 Iteration time: 0.368s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0863 Mean episode length: 1371.3 Actor loss: -9.4627 Critic loss: +0.0040 World model loss: +0.8323 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2933 Episode_Reward/track_angular_velocity: +51.8362 Episode_Reward/upright: +26.5728 Episode_Reward/leg_posture: -2.7791 Episode_Reward/standstill: -6.1921 Episode_Reward/action_rate_l2: -0.6171 Episode_Reward/joint_torques_l2: -0.1494 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.1064 Episode_Reward/rolling_leg_motion: -0.7543 Episode_Reward/obstacle_progress: +2.7605 Episode_Reward/wheel_body_intrusion: -0.2583 Episode_Reward/pair_body_intrusion: -0.3097 Curriculum/terrain_level_mean: +2.5771 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4892 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.4676 Episode_Termination/illegal_nonwheel_contact: 0.0216 ============================================================================== ============================================================================== Learning iteration 14900/50000 Total steps: 61,030,400 Steps per second: 11,136 Iteration time: 0.368s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1272.0 Actor loss: -9.4647 Critic loss: +0.0050 World model loss: +0.8870 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.7311 Episode_Reward/track_angular_velocity: +47.9932 Episode_Reward/upright: +24.6377 Episode_Reward/leg_posture: -2.6218 Episode_Reward/standstill: -5.7248 Episode_Reward/action_rate_l2: -0.5738 Episode_Reward/joint_torques_l2: -0.1382 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0174 Episode_Reward/unnecessary_wheel_lift: -0.1015 Episode_Reward/rolling_leg_motion: -0.7067 Episode_Reward/obstacle_progress: +2.4945 Episode_Reward/wheel_body_intrusion: -0.2607 Episode_Reward/pair_body_intrusion: -0.2818 Curriculum/terrain_level_mean: +2.5845 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0973 Metrics/base_velocity/error_vel_yaw: 0.1381 Episode_Termination/time_out: 0.4069 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.5448 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== ============================================================================== Learning iteration 14950/50000 Total steps: 61,235,200 Steps per second: 11,147 Iteration time: 0.367s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1259.6 Actor loss: -9.4437 Critic loss: +0.0048 World model loss: +0.8135 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.2921 Episode_Reward/track_angular_velocity: +47.5981 Episode_Reward/upright: +24.4137 Episode_Reward/leg_posture: -2.5820 Episode_Reward/standstill: -5.6916 Episode_Reward/action_rate_l2: -0.5555 Episode_Reward/joint_torques_l2: -0.1371 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0197 Episode_Reward/unnecessary_wheel_lift: -0.1199 Episode_Reward/rolling_leg_motion: -0.7090 Episode_Reward/obstacle_progress: +2.4380 Episode_Reward/wheel_body_intrusion: -0.2471 Episode_Reward/pair_body_intrusion: -0.2548 Curriculum/terrain_level_mean: +2.5933 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0974 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4126 Episode_Termination/fell_over: 0.0350 Episode_Termination/course_exit: 0.5315 Episode_Termination/illegal_nonwheel_contact: 0.0490 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2205] ============================================================================== Learning iteration 15000/50000 Total steps: 61,440,000 Steps per second: 11,157 Iteration time: 0.367s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0861 Mean episode length: 1279.5 Actor loss: -9.4857 Critic loss: +0.0027 World model loss: +0.7895 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.0061 Episode_Reward/track_angular_velocity: +48.3497 Episode_Reward/upright: +24.8104 Episode_Reward/leg_posture: -2.5939 Episode_Reward/standstill: -5.8107 Episode_Reward/action_rate_l2: -0.5698 Episode_Reward/joint_torques_l2: -0.1378 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0225 Episode_Reward/unnecessary_wheel_lift: -0.1127 Episode_Reward/rolling_leg_motion: -0.7212 Episode_Reward/obstacle_progress: +2.4745 Episode_Reward/wheel_body_intrusion: -0.2516 Episode_Reward/pair_body_intrusion: -0.2561 Curriculum/terrain_level_mean: +2.6021 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4375 Episode_Termination/fell_over: 0.0250 Episode_Termination/course_exit: 0.5062 Episode_Termination/illegal_nonwheel_contact: 0.0437 ============================================================================== ============================================================================== Learning iteration 15050/50000 Total steps: 61,644,800 Steps per second: 11,168 Iteration time: 0.367s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1344.8 Actor loss: -9.4798 Critic loss: +0.0030 World model loss: +0.8664 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2128 Episode_Reward/track_angular_velocity: +50.8374 Episode_Reward/upright: +26.0524 Episode_Reward/leg_posture: -2.7656 Episode_Reward/standstill: -5.9485 Episode_Reward/action_rate_l2: -0.5963 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0131 Episode_Reward/unnecessary_wheel_lift: -0.1167 Episode_Reward/rolling_leg_motion: -0.7460 Episode_Reward/obstacle_progress: +2.5891 Episode_Reward/wheel_body_intrusion: -0.2583 Episode_Reward/pair_body_intrusion: -0.3317 Curriculum/terrain_level_mean: +2.6106 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0979 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4400 Episode_Termination/fell_over: 0.0133 Episode_Termination/course_exit: 0.5267 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[909] ============================================================================== Learning iteration 15100/50000 Total steps: 61,849,600 Steps per second: 11,178 Iteration time: 0.366s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1368.4 Actor loss: -9.4677 Critic loss: +0.0046 World model loss: +0.8013 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0419 Episode_Reward/track_angular_velocity: +51.7315 Episode_Reward/upright: +26.5078 Episode_Reward/leg_posture: -2.8416 Episode_Reward/standstill: -6.0225 Episode_Reward/action_rate_l2: -0.5885 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.1131 Episode_Reward/rolling_leg_motion: -0.7565 Episode_Reward/obstacle_progress: +2.6919 Episode_Reward/wheel_body_intrusion: -0.2943 Episode_Reward/pair_body_intrusion: -0.3783 Curriculum/terrain_level_mean: +2.6157 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0975 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5106 Episode_Termination/fell_over: 0.0496 Episode_Termination/course_exit: 0.4184 Episode_Termination/illegal_nonwheel_contact: 0.0355 ============================================================================== ============================================================================== Learning iteration 15150/50000 Total steps: 62,054,400 Steps per second: 11,189 Iteration time: 0.366s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1330.7 Actor loss: -9.4523 Critic loss: +0.0034 World model loss: +0.8548 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7054 Episode_Reward/track_angular_velocity: +50.2469 Episode_Reward/upright: +25.7494 Episode_Reward/leg_posture: -2.7758 Episode_Reward/standstill: -5.8372 Episode_Reward/action_rate_l2: -0.5917 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.1063 Episode_Reward/rolling_leg_motion: -0.7489 Episode_Reward/obstacle_progress: +2.6393 Episode_Reward/wheel_body_intrusion: -0.2779 Episode_Reward/pair_body_intrusion: -0.3105 Curriculum/terrain_level_mean: +2.6235 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0977 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.5357 Episode_Termination/fell_over: 0.0214 Episode_Termination/course_exit: 0.4214 Episode_Termination/illegal_nonwheel_contact: 0.0214 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[631] ============================================================================== Learning iteration 15200/50000 Total steps: 62,259,200 Steps per second: 11,199 Iteration time: 0.366s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0858 Mean episode length: 1368.8 Actor loss: -9.4603 Critic loss: +0.0054 World model loss: +0.8149 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9610 Episode_Reward/track_angular_velocity: +51.6707 Episode_Reward/upright: +26.4771 Episode_Reward/leg_posture: -2.8476 Episode_Reward/standstill: -6.2446 Episode_Reward/action_rate_l2: -0.6101 Episode_Reward/joint_torques_l2: -0.1482 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.1388 Episode_Reward/rolling_leg_motion: -0.7842 Episode_Reward/obstacle_progress: +2.7088 Episode_Reward/wheel_body_intrusion: -0.2814 Episode_Reward/pair_body_intrusion: -0.3098 Curriculum/terrain_level_mean: +2.6318 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0987 Metrics/base_velocity/error_vel_yaw: 0.1375 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0373 Episode_Termination/course_exit: 0.4596 Episode_Termination/illegal_nonwheel_contact: 0.0311 ============================================================================== ============================================================================== Learning iteration 15250/50000 Total steps: 62,464,000 Steps per second: 11,210 Iteration time: 0.365s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1345.0 Actor loss: -9.4748 Critic loss: +0.0076 World model loss: +0.8131 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3585 Episode_Reward/track_angular_velocity: +50.8438 Episode_Reward/upright: +26.0451 Episode_Reward/leg_posture: -2.7503 Episode_Reward/standstill: -6.0565 Episode_Reward/action_rate_l2: -0.5946 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.1260 Episode_Reward/rolling_leg_motion: -0.7498 Episode_Reward/obstacle_progress: +2.7159 Episode_Reward/wheel_body_intrusion: -0.2605 Episode_Reward/pair_body_intrusion: -0.2921 Curriculum/terrain_level_mean: +2.6396 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0964 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4710 Episode_Termination/fell_over: 0.0387 Episode_Termination/course_exit: 0.4774 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== ============================================================================== Learning iteration 15300/50000 Total steps: 62,668,800 Steps per second: 11,220 Iteration time: 0.365s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0863 Mean episode length: 1325.4 Actor loss: -9.4958 Critic loss: +0.0030 World model loss: +0.8311 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7907 Episode_Reward/track_angular_velocity: +50.1600 Episode_Reward/upright: +25.6941 Episode_Reward/leg_posture: -2.7356 Episode_Reward/standstill: -6.0716 Episode_Reward/action_rate_l2: -0.5741 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.1063 Episode_Reward/rolling_leg_motion: -0.7299 Episode_Reward/obstacle_progress: +2.6839 Episode_Reward/wheel_body_intrusion: -0.2533 Episode_Reward/pair_body_intrusion: -0.2887 Curriculum/terrain_level_mean: +2.6462 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0952 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4565 Episode_Termination/fell_over: 0.0435 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0145 ============================================================================== ============================================================================== Learning iteration 15350/50000 Total steps: 62,873,600 Steps per second: 11,231 Iteration time: 0.365s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0863 Mean episode length: 1365.6 Actor loss: -9.4446 Critic loss: +0.0063 World model loss: +0.8516 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1736 Episode_Reward/track_angular_velocity: +51.6729 Episode_Reward/upright: +26.4581 Episode_Reward/leg_posture: -2.8322 Episode_Reward/standstill: -6.2371 Episode_Reward/action_rate_l2: -0.5955 Episode_Reward/joint_torques_l2: -0.1475 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0053 Episode_Reward/unnecessary_wheel_lift: -0.1046 Episode_Reward/rolling_leg_motion: -0.7542 Episode_Reward/obstacle_progress: +2.7526 Episode_Reward/wheel_body_intrusion: -0.2751 Episode_Reward/pair_body_intrusion: -0.2856 Curriculum/terrain_level_mean: +2.6523 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0955 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4786 Episode_Termination/fell_over: 0.0071 Episode_Termination/course_exit: 0.5071 Episode_Termination/illegal_nonwheel_contact: 0.0071 ============================================================================== ============================================================================== Learning iteration 15400/50000 Total steps: 63,078,400 Steps per second: 11,241 Iteration time: 0.364s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0860 Mean episode length: 1449.9 Actor loss: -9.4587 Critic loss: +0.0027 World model loss: +0.8336 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0895 Episode_Reward/track_angular_velocity: +54.8123 Episode_Reward/upright: +28.0543 Episode_Reward/leg_posture: -2.9895 Episode_Reward/standstill: -6.9345 Episode_Reward/action_rate_l2: -0.6279 Episode_Reward/joint_torques_l2: -0.1561 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0057 Episode_Reward/unnecessary_wheel_lift: -0.1308 Episode_Reward/rolling_leg_motion: -0.8116 Episode_Reward/obstacle_progress: +2.9457 Episode_Reward/wheel_body_intrusion: -0.2821 Episode_Reward/pair_body_intrusion: -0.2935 Curriculum/terrain_level_mean: +2.6580 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5683 Episode_Termination/fell_over: 0.0144 Episode_Termination/course_exit: 0.4173 Episode_Termination/illegal_nonwheel_contact: 0.0072 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1862] ============================================================================== Learning iteration 15450/50000 Total steps: 63,283,200 Steps per second: 11,251 Iteration time: 0.364s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0860 Mean episode length: 1338.0 Actor loss: -9.4691 Critic loss: +0.0036 World model loss: +0.8014 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0617 Episode_Reward/track_angular_velocity: +50.5909 Episode_Reward/upright: +25.8704 Episode_Reward/leg_posture: -2.7813 Episode_Reward/standstill: -6.2869 Episode_Reward/action_rate_l2: -0.5793 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1230 Episode_Reward/rolling_leg_motion: -0.7431 Episode_Reward/obstacle_progress: +2.7256 Episode_Reward/wheel_body_intrusion: -0.2646 Episode_Reward/pair_body_intrusion: -0.2716 Curriculum/terrain_level_mean: +2.6641 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0294 Episode_Termination/course_exit: 0.4485 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 15500/50000 Total steps: 63,488,000 Steps per second: 11,262 Iteration time: 0.364s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0860 Mean episode length: 1370.7 Actor loss: -9.4661 Critic loss: +0.0041 World model loss: +0.8204 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2271 Episode_Reward/track_angular_velocity: +51.7726 Episode_Reward/upright: +26.5111 Episode_Reward/leg_posture: -2.8867 Episode_Reward/standstill: -6.3082 Episode_Reward/action_rate_l2: -0.5932 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.1230 Episode_Reward/rolling_leg_motion: -0.7874 Episode_Reward/obstacle_progress: +2.7007 Episode_Reward/wheel_body_intrusion: -0.2655 Episode_Reward/pair_body_intrusion: -0.2632 Curriculum/terrain_level_mean: +2.6694 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.5036 Episode_Termination/fell_over: 0.0288 Episode_Termination/course_exit: 0.4460 Episode_Termination/illegal_nonwheel_contact: 0.0504 ============================================================================== ============================================================================== Learning iteration 15550/50000 Total steps: 63,692,800 Steps per second: 11,271 Iteration time: 0.363s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0862 Mean episode length: 1373.6 Actor loss: -9.4637 Critic loss: +0.0052 World model loss: +0.8403 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3856 Episode_Reward/track_angular_velocity: +51.9362 Episode_Reward/upright: +26.5803 Episode_Reward/leg_posture: -2.8517 Episode_Reward/standstill: -6.0888 Episode_Reward/action_rate_l2: -0.6115 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.1106 Episode_Reward/rolling_leg_motion: -0.7738 Episode_Reward/obstacle_progress: +2.6724 Episode_Reward/wheel_body_intrusion: -0.2703 Episode_Reward/pair_body_intrusion: -0.2798 Curriculum/terrain_level_mean: +2.6785 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4785 Episode_Termination/fell_over: 0.0184 Episode_Termination/course_exit: 0.4847 Episode_Termination/illegal_nonwheel_contact: 0.0245 ============================================================================== ============================================================================== Learning iteration 15600/50000 Total steps: 63,897,600 Steps per second: 11,281 Iteration time: 0.363s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0859 Mean episode length: 1343.5 Actor loss: -9.4840 Critic loss: +0.0028 World model loss: +0.7809 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3448 Episode_Reward/track_angular_velocity: +50.7509 Episode_Reward/upright: +26.0038 Episode_Reward/leg_posture: -2.7851 Episode_Reward/standstill: -6.4065 Episode_Reward/action_rate_l2: -0.6078 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0131 Episode_Reward/unnecessary_wheel_lift: -0.1251 Episode_Reward/rolling_leg_motion: -0.7600 Episode_Reward/obstacle_progress: +2.7200 Episode_Reward/wheel_body_intrusion: -0.2643 Episode_Reward/pair_body_intrusion: -0.2839 Curriculum/terrain_level_mean: +2.6882 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0961 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4611 Episode_Termination/fell_over: 0.0111 Episode_Termination/course_exit: 0.5111 Episode_Termination/illegal_nonwheel_contact: 0.0167 ============================================================================== ============================================================================== Learning iteration 15650/50000 Total steps: 64,102,400 Steps per second: 11,291 Iteration time: 0.363s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0859 Mean episode length: 1314.0 Actor loss: -9.4701 Critic loss: +0.0102 World model loss: +0.8220 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2902 Episode_Reward/track_angular_velocity: +49.6389 Episode_Reward/upright: +25.4010 Episode_Reward/leg_posture: -2.7485 Episode_Reward/standstill: -6.2142 Episode_Reward/action_rate_l2: -0.5886 Episode_Reward/joint_torques_l2: -0.1402 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.1438 Episode_Reward/rolling_leg_motion: -0.7279 Episode_Reward/obstacle_progress: +2.6800 Episode_Reward/wheel_body_intrusion: -0.2627 Episode_Reward/pair_body_intrusion: -0.3071 Curriculum/terrain_level_mean: +2.6938 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.4730 Episode_Termination/illegal_nonwheel_contact: 0.0608 ============================================================================== ============================================================================== Learning iteration 15700/50000 Total steps: 64,307,200 Steps per second: 11,302 Iteration time: 0.362s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1386.0 Actor loss: -9.4785 Critic loss: +0.0032 World model loss: +0.8220 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8020 Episode_Reward/track_angular_velocity: +52.4332 Episode_Reward/upright: +26.8148 Episode_Reward/leg_posture: -2.8708 Episode_Reward/standstill: -6.1829 Episode_Reward/action_rate_l2: -0.6145 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1272 Episode_Reward/rolling_leg_motion: -0.7816 Episode_Reward/obstacle_progress: +2.7912 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.3603 Curriculum/terrain_level_mean: +2.7014 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4899 Episode_Termination/fell_over: 0.0268 Episode_Termination/course_exit: 0.4631 Episode_Termination/illegal_nonwheel_contact: 0.0268 ============================================================================== ============================================================================== Learning iteration 15750/50000 Total steps: 64,512,000 Steps per second: 11,312 Iteration time: 0.362s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0857 Mean episode length: 1357.9 Actor loss: -9.4688 Critic loss: +0.0045 World model loss: +0.7805 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5628 Episode_Reward/track_angular_velocity: +51.3033 Episode_Reward/upright: +26.2622 Episode_Reward/leg_posture: -2.8582 Episode_Reward/standstill: -6.3537 Episode_Reward/action_rate_l2: -0.5878 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0239 Episode_Reward/unnecessary_wheel_lift: -0.1209 Episode_Reward/rolling_leg_motion: -0.7518 Episode_Reward/obstacle_progress: +2.6712 Episode_Reward/wheel_body_intrusion: -0.2933 Episode_Reward/pair_body_intrusion: -0.3215 Curriculum/terrain_level_mean: +2.7073 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0984 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5145 Episode_Termination/fell_over: 0.0231 Episode_Termination/course_exit: 0.4335 Episode_Termination/illegal_nonwheel_contact: 0.0405 ============================================================================== ============================================================================== Learning iteration 15800/50000 Total steps: 64,716,800 Steps per second: 11,322 Iteration time: 0.362s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0858 Mean episode length: 1351.3 Actor loss: -9.4616 Critic loss: +0.0035 World model loss: +0.8144 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5114 Episode_Reward/track_angular_velocity: +51.0393 Episode_Reward/upright: +26.1057 Episode_Reward/leg_posture: -2.8443 Episode_Reward/standstill: -6.3736 Episode_Reward/action_rate_l2: -0.5806 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1160 Episode_Reward/rolling_leg_motion: -0.7444 Episode_Reward/obstacle_progress: +2.6709 Episode_Reward/wheel_body_intrusion: -0.2841 Episode_Reward/pair_body_intrusion: -0.3477 Curriculum/terrain_level_mean: +2.7131 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4362 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.5235 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 15850/50000 Total steps: 64,921,600 Steps per second: 11,332 Iteration time: 0.361s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1336.5 Actor loss: -9.4545 Critic loss: +0.0030 World model loss: +0.7897 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9838 Episode_Reward/track_angular_velocity: +50.4864 Episode_Reward/upright: +25.8126 Episode_Reward/leg_posture: -2.8687 Episode_Reward/standstill: -6.4054 Episode_Reward/action_rate_l2: -0.5750 Episode_Reward/joint_torques_l2: -0.1455 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1356 Episode_Reward/rolling_leg_motion: -0.7575 Episode_Reward/obstacle_progress: +2.6844 Episode_Reward/wheel_body_intrusion: -0.2791 Episode_Reward/pair_body_intrusion: -0.3489 Curriculum/terrain_level_mean: +2.7207 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4620 Episode_Termination/fell_over: 0.0253 Episode_Termination/course_exit: 0.4937 Episode_Termination/illegal_nonwheel_contact: 0.0253 ============================================================================== ============================================================================== Learning iteration 15900/50000 Total steps: 65,126,400 Steps per second: 11,342 Iteration time: 0.361s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1396.8 Actor loss: -9.4546 Critic loss: +0.0027 World model loss: +0.7713 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1766 Episode_Reward/track_angular_velocity: +52.8056 Episode_Reward/upright: +27.0174 Episode_Reward/leg_posture: -2.9277 Episode_Reward/standstill: -6.5312 Episode_Reward/action_rate_l2: -0.5997 Episode_Reward/joint_torques_l2: -0.1518 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0273 Episode_Reward/unnecessary_wheel_lift: -0.1339 Episode_Reward/rolling_leg_motion: -0.7644 Episode_Reward/obstacle_progress: +2.8623 Episode_Reward/wheel_body_intrusion: -0.2855 Episode_Reward/pair_body_intrusion: -0.3417 Curriculum/terrain_level_mean: +2.7256 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0965 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5294 Episode_Termination/fell_over: 0.0458 Episode_Termination/course_exit: 0.4052 Episode_Termination/illegal_nonwheel_contact: 0.0458 ============================================================================== ============================================================================== Learning iteration 15950/50000 Total steps: 65,331,200 Steps per second: 11,352 Iteration time: 0.361s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1315.4 Actor loss: -9.4756 Critic loss: +0.0032 World model loss: +0.7728 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3549 Episode_Reward/track_angular_velocity: +49.7938 Episode_Reward/upright: +25.4456 Episode_Reward/leg_posture: -2.7226 Episode_Reward/standstill: -6.3443 Episode_Reward/action_rate_l2: -0.5621 Episode_Reward/joint_torques_l2: -0.1413 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0154 Episode_Reward/unnecessary_wheel_lift: -0.1266 Episode_Reward/rolling_leg_motion: -0.7108 Episode_Reward/obstacle_progress: +2.6951 Episode_Reward/wheel_body_intrusion: -0.2636 Episode_Reward/pair_body_intrusion: -0.3327 Curriculum/terrain_level_mean: +2.7322 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.3836 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.5753 Episode_Termination/illegal_nonwheel_contact: 0.0342 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[446] ============================================================================== Learning iteration 16000/50000 Total steps: 65,536,000 Steps per second: 11,362 Iteration time: 0.361s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0856 Mean episode length: 1386.6 Actor loss: -9.4904 Critic loss: +0.0027 World model loss: +0.8370 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8119 Episode_Reward/track_angular_velocity: +52.3810 Episode_Reward/upright: +26.8159 Episode_Reward/leg_posture: -2.9017 Episode_Reward/standstill: -7.0339 Episode_Reward/action_rate_l2: -0.5918 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.1414 Episode_Reward/rolling_leg_motion: -0.7662 Episode_Reward/obstacle_progress: +2.8531 Episode_Reward/wheel_body_intrusion: -0.2774 Episode_Reward/pair_body_intrusion: -0.3457 Curriculum/terrain_level_mean: +2.7368 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5203 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.4122 Episode_Termination/illegal_nonwheel_contact: 0.0676 ============================================================================== ============================================================================== Learning iteration 16050/50000 Total steps: 65,740,800 Steps per second: 11,371 Iteration time: 0.360s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0857 Mean episode length: 1413.8 Actor loss: -9.4499 Critic loss: +0.0028 World model loss: +0.8072 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6533 Episode_Reward/track_angular_velocity: +53.4314 Episode_Reward/upright: +27.3317 Episode_Reward/leg_posture: -2.9704 Episode_Reward/standstill: -6.7647 Episode_Reward/action_rate_l2: -0.6157 Episode_Reward/joint_torques_l2: -0.1541 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0193 Episode_Reward/unnecessary_wheel_lift: -0.1499 Episode_Reward/rolling_leg_motion: -0.7806 Episode_Reward/obstacle_progress: +2.8742 Episode_Reward/wheel_body_intrusion: -0.2961 Episode_Reward/pair_body_intrusion: -0.3835 Curriculum/terrain_level_mean: +2.7434 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0974 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.5350 Episode_Termination/fell_over: 0.0318 Episode_Termination/course_exit: 0.4204 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== Warning: EPA horizon = 24 isn't large enough. Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 16100/50000 Total steps: 65,945,600 Steps per second: 11,381 Iteration time: 0.360s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0857 Mean episode length: 1324.4 Actor loss: -9.4618 Critic loss: +0.0053 World model loss: +0.8246 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4759 Episode_Reward/track_angular_velocity: +50.0059 Episode_Reward/upright: +25.5699 Episode_Reward/leg_posture: -2.7201 Episode_Reward/standstill: -6.4229 Episode_Reward/action_rate_l2: -0.5629 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0310 Episode_Reward/unnecessary_wheel_lift: -0.1364 Episode_Reward/rolling_leg_motion: -0.7149 Episode_Reward/obstacle_progress: +2.7686 Episode_Reward/wheel_body_intrusion: -0.2742 Episode_Reward/pair_body_intrusion: -0.3024 Curriculum/terrain_level_mean: +2.7478 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0977 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4540 Episode_Termination/fell_over: 0.0552 Episode_Termination/course_exit: 0.4724 Episode_Termination/illegal_nonwheel_contact: 0.0429 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[873] ============================================================================== Learning iteration 16150/50000 Total steps: 66,150,400 Steps per second: 11,391 Iteration time: 0.360s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1451.1 Actor loss: -9.4707 Critic loss: +0.0074 World model loss: +0.7973 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0072 Episode_Reward/track_angular_velocity: +54.8858 Episode_Reward/upright: +28.0491 Episode_Reward/leg_posture: -3.0029 Episode_Reward/standstill: -6.6945 Episode_Reward/action_rate_l2: -0.6161 Episode_Reward/joint_torques_l2: -0.1574 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0028 Episode_Reward/unnecessary_wheel_lift: -0.1185 Episode_Reward/rolling_leg_motion: -0.7570 Episode_Reward/obstacle_progress: +3.0724 Episode_Reward/wheel_body_intrusion: -0.2788 Episode_Reward/pair_body_intrusion: -0.3383 Curriculum/terrain_level_mean: +2.7537 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.6164 Episode_Termination/fell_over: 0.0063 Episode_Termination/course_exit: 0.3774 Episode_Termination/illegal_nonwheel_contact: 0.0063 ============================================================================== ============================================================================== Learning iteration 16200/50000 Total steps: 66,355,200 Steps per second: 11,401 Iteration time: 0.359s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0858 Mean episode length: 1375.1 Actor loss: -9.4375 Critic loss: +0.0042 World model loss: +0.7837 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2227 Episode_Reward/track_angular_velocity: +51.9484 Episode_Reward/upright: +26.5763 Episode_Reward/leg_posture: -2.8411 Episode_Reward/standstill: -6.6206 Episode_Reward/action_rate_l2: -0.5986 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0215 Episode_Reward/unnecessary_wheel_lift: -0.1248 Episode_Reward/rolling_leg_motion: -0.7510 Episode_Reward/obstacle_progress: +2.8823 Episode_Reward/wheel_body_intrusion: -0.2746 Episode_Reward/pair_body_intrusion: -0.3286 Curriculum/terrain_level_mean: +2.7607 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0980 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4852 Episode_Termination/fell_over: 0.0414 Episode_Termination/course_exit: 0.4675 Episode_Termination/illegal_nonwheel_contact: 0.0178 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_016200.mp4 ============================================================================== Learning iteration 16250/50000 Total steps: 66,560,000 Steps per second: 11,399 Iteration time: 0.359s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0859 Mean episode length: 1322.9 Actor loss: -9.4310 Critic loss: +0.0074 World model loss: +0.8031 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3942 Episode_Reward/track_angular_velocity: +49.9928 Episode_Reward/upright: +25.5815 Episode_Reward/leg_posture: -2.7225 Episode_Reward/standstill: -6.1558 Episode_Reward/action_rate_l2: -0.5738 Episode_Reward/joint_torques_l2: -0.1425 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.1157 Episode_Reward/rolling_leg_motion: -0.7198 Episode_Reward/obstacle_progress: +2.7469 Episode_Reward/wheel_body_intrusion: -0.2515 Episode_Reward/pair_body_intrusion: -0.3171 Curriculum/terrain_level_mean: +2.7668 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0977 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.5066 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.4408 Episode_Termination/illegal_nonwheel_contact: 0.0329 ============================================================================== ============================================================================== Learning iteration 16300/50000 Total steps: 66,764,800 Steps per second: 11,409 Iteration time: 0.359s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0857 Mean episode length: 1293.0 Actor loss: -9.4562 Critic loss: +0.0034 World model loss: +0.7833 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.3476 Episode_Reward/track_angular_velocity: +48.8038 Episode_Reward/upright: +24.9849 Episode_Reward/leg_posture: -2.6813 Episode_Reward/standstill: -6.1654 Episode_Reward/action_rate_l2: -0.5510 Episode_Reward/joint_torques_l2: -0.1391 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0278 Episode_Reward/unnecessary_wheel_lift: -0.1312 Episode_Reward/rolling_leg_motion: -0.7109 Episode_Reward/obstacle_progress: +2.6869 Episode_Reward/wheel_body_intrusion: -0.2486 Episode_Reward/pair_body_intrusion: -0.3290 Curriculum/terrain_level_mean: +2.7739 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0979 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.4713 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4586 Episode_Termination/illegal_nonwheel_contact: 0.0573 ============================================================================== ============================================================================== Learning iteration 16350/50000 Total steps: 66,969,600 Steps per second: 11,418 Iteration time: 0.359s Collection time: 0.075s Learning time: 0.193s Mean reward: +0.0859 Mean episode length: 1384.4 Actor loss: -9.4547 Critic loss: +0.0069 World model loss: +0.7731 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5993 Episode_Reward/track_angular_velocity: +52.3616 Episode_Reward/upright: +26.7659 Episode_Reward/leg_posture: -2.9034 Episode_Reward/standstill: -6.4136 Episode_Reward/action_rate_l2: -0.5902 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.1186 Episode_Reward/rolling_leg_motion: -0.7673 Episode_Reward/obstacle_progress: +2.7908 Episode_Reward/wheel_body_intrusion: -0.2830 Episode_Reward/pair_body_intrusion: -0.3321 Curriculum/terrain_level_mean: +2.7808 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5524 Episode_Termination/fell_over: 0.0140 Episode_Termination/course_exit: 0.4126 Episode_Termination/illegal_nonwheel_contact: 0.0280 ============================================================================== ============================================================================== Learning iteration 16400/50000 Total steps: 67,174,400 Steps per second: 11,427 Iteration time: 0.358s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1367.3 Actor loss: -9.4365 Critic loss: +0.0053 World model loss: +0.8191 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9561 Episode_Reward/track_angular_velocity: +51.7387 Episode_Reward/upright: +26.4051 Episode_Reward/leg_posture: -2.8269 Episode_Reward/standstill: -6.2923 Episode_Reward/action_rate_l2: -0.5850 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1243 Episode_Reward/rolling_leg_motion: -0.7484 Episode_Reward/obstacle_progress: +2.8054 Episode_Reward/wheel_body_intrusion: -0.2697 Episode_Reward/pair_body_intrusion: -0.2828 Curriculum/terrain_level_mean: +2.7849 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0978 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4925 Episode_Termination/fell_over: 0.0373 Episode_Termination/course_exit: 0.4478 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2421] ============================================================================== Learning iteration 16450/50000 Total steps: 67,379,200 Steps per second: 11,436 Iteration time: 0.358s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0858 Mean episode length: 1369.5 Actor loss: -9.4301 Critic loss: +0.0029 World model loss: +0.7401 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1420 Episode_Reward/track_angular_velocity: +51.8286 Episode_Reward/upright: +26.4676 Episode_Reward/leg_posture: -2.8518 Episode_Reward/standstill: -6.6493 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1477 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.1105 Episode_Reward/rolling_leg_motion: -0.7544 Episode_Reward/obstacle_progress: +2.7799 Episode_Reward/wheel_body_intrusion: -0.2789 Episode_Reward/pair_body_intrusion: -0.3006 Curriculum/terrain_level_mean: +2.7903 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0411 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0274 ============================================================================== ============================================================================== Learning iteration 16500/50000 Total steps: 67,584,000 Steps per second: 11,446 Iteration time: 0.358s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1363.2 Actor loss: -9.4236 Critic loss: +0.0037 World model loss: +0.8047 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9801 Episode_Reward/track_angular_velocity: +51.6183 Episode_Reward/upright: +26.3205 Episode_Reward/leg_posture: -2.9298 Episode_Reward/standstill: -6.2402 Episode_Reward/action_rate_l2: -0.5756 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0266 Episode_Reward/unnecessary_wheel_lift: -0.1082 Episode_Reward/rolling_leg_motion: -0.7704 Episode_Reward/obstacle_progress: +2.7016 Episode_Reward/wheel_body_intrusion: -0.2779 Episode_Reward/pair_body_intrusion: -0.3235 Curriculum/terrain_level_mean: +2.7961 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0960 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.5342 Episode_Termination/fell_over: 0.0616 Episode_Termination/course_exit: 0.3973 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 16550/50000 Total steps: 67,788,800 Steps per second: 11,455 Iteration time: 0.358s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0858 Mean episode length: 1320.6 Actor loss: -9.4429 Critic loss: +0.0065 World model loss: +0.7656 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4373 Episode_Reward/track_angular_velocity: +49.9490 Episode_Reward/upright: +25.4978 Episode_Reward/leg_posture: -2.8609 Episode_Reward/standstill: -6.1875 Episode_Reward/action_rate_l2: -0.5538 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.1173 Episode_Reward/rolling_leg_motion: -0.7459 Episode_Reward/obstacle_progress: +2.5623 Episode_Reward/wheel_body_intrusion: -0.2607 Episode_Reward/pair_body_intrusion: -0.2824 Curriculum/terrain_level_mean: +2.8037 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0964 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4565 Episode_Termination/fell_over: 0.0290 Episode_Termination/course_exit: 0.4855 Episode_Termination/illegal_nonwheel_contact: 0.0290 ============================================================================== ============================================================================== Learning iteration 16600/50000 Total steps: 67,993,600 Steps per second: 11,465 Iteration time: 0.357s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0855 Mean episode length: 1364.6 Actor loss: -9.4239 Critic loss: +0.0060 World model loss: +0.7922 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9580 Episode_Reward/track_angular_velocity: +51.5714 Episode_Reward/upright: +26.3595 Episode_Reward/leg_posture: -2.9273 Episode_Reward/standstill: -6.7133 Episode_Reward/action_rate_l2: -0.5741 Episode_Reward/joint_torques_l2: -0.1465 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.1146 Episode_Reward/rolling_leg_motion: -0.7513 Episode_Reward/obstacle_progress: +2.6850 Episode_Reward/wheel_body_intrusion: -0.3113 Episode_Reward/pair_body_intrusion: -0.3344 Curriculum/terrain_level_mean: +2.8105 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0966 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4904 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.4777 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 16650/50000 Total steps: 68,198,400 Steps per second: 11,474 Iteration time: 0.357s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0857 Mean episode length: 1314.4 Actor loss: -9.4586 Critic loss: +0.0031 World model loss: +0.8067 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2841 Episode_Reward/track_angular_velocity: +49.7358 Episode_Reward/upright: +25.3926 Episode_Reward/leg_posture: -2.8264 Episode_Reward/standstill: -6.2835 Episode_Reward/action_rate_l2: -0.5625 Episode_Reward/joint_torques_l2: -0.1394 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0154 Episode_Reward/unnecessary_wheel_lift: -0.1084 Episode_Reward/rolling_leg_motion: -0.7370 Episode_Reward/obstacle_progress: +2.5572 Episode_Reward/wheel_body_intrusion: -0.2600 Episode_Reward/pair_body_intrusion: -0.3358 Curriculum/terrain_level_mean: +2.8154 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0959 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4539 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.5035 Episode_Termination/illegal_nonwheel_contact: 0.0284 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[549] ============================================================================== Learning iteration 16700/50000 Total steps: 68,403,200 Steps per second: 11,484 Iteration time: 0.357s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0855 Mean episode length: 1314.6 Actor loss: -9.4397 Critic loss: +0.0061 World model loss: +0.7974 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2383 Episode_Reward/track_angular_velocity: +49.6844 Episode_Reward/upright: +25.3338 Episode_Reward/leg_posture: -2.7765 Episode_Reward/standstill: -6.4557 Episode_Reward/action_rate_l2: -0.5760 Episode_Reward/joint_torques_l2: -0.1413 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0180 Episode_Reward/unnecessary_wheel_lift: -0.1205 Episode_Reward/rolling_leg_motion: -0.7325 Episode_Reward/obstacle_progress: +2.5789 Episode_Reward/wheel_body_intrusion: -0.2767 Episode_Reward/pair_body_intrusion: -0.3249 Curriculum/terrain_level_mean: +2.8196 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4437 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.5070 Episode_Termination/illegal_nonwheel_contact: 0.0423 ============================================================================== ============================================================================== Learning iteration 16750/50000 Total steps: 68,608,000 Steps per second: 11,493 Iteration time: 0.356s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0857 Mean episode length: 1340.2 Actor loss: -9.4718 Critic loss: +0.0035 World model loss: +0.7961 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2386 Episode_Reward/track_angular_velocity: +50.6196 Episode_Reward/upright: +25.7885 Episode_Reward/leg_posture: -2.8818 Episode_Reward/standstill: -6.3156 Episode_Reward/action_rate_l2: -0.5936 Episode_Reward/joint_torques_l2: -0.1467 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1309 Episode_Reward/rolling_leg_motion: -0.7344 Episode_Reward/obstacle_progress: +2.6706 Episode_Reward/wheel_body_intrusion: -0.3149 Episode_Reward/pair_body_intrusion: -0.3464 Curriculum/terrain_level_mean: +2.8276 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4743 Episode_Termination/fell_over: 0.0229 Episode_Termination/course_exit: 0.4857 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== ============================================================================== Learning iteration 16800/50000 Total steps: 68,812,800 Steps per second: 11,502 Iteration time: 0.356s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1361.8 Actor loss: -9.4243 Critic loss: +0.0040 World model loss: +0.7506 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1056 Episode_Reward/track_angular_velocity: +51.5195 Episode_Reward/upright: +26.2294 Episode_Reward/leg_posture: -2.8680 Episode_Reward/standstill: -6.5628 Episode_Reward/action_rate_l2: -0.5962 Episode_Reward/joint_torques_l2: -0.1493 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0053 Episode_Reward/unnecessary_wheel_lift: -0.1017 Episode_Reward/rolling_leg_motion: -0.7257 Episode_Reward/obstacle_progress: +2.8660 Episode_Reward/wheel_body_intrusion: -0.3042 Episode_Reward/pair_body_intrusion: -0.3559 Curriculum/terrain_level_mean: +2.8345 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0952 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4899 Episode_Termination/fell_over: 0.0067 Episode_Termination/course_exit: 0.4966 Episode_Termination/illegal_nonwheel_contact: 0.0067 ============================================================================== ============================================================================== Learning iteration 16850/50000 Total steps: 69,017,600 Steps per second: 11,511 Iteration time: 0.356s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1358.9 Actor loss: -9.4653 Critic loss: +0.0032 World model loss: +0.7636 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7607 Episode_Reward/track_angular_velocity: +51.3554 Episode_Reward/upright: +26.1470 Episode_Reward/leg_posture: -2.8267 Episode_Reward/standstill: -6.6885 Episode_Reward/action_rate_l2: -0.6098 Episode_Reward/joint_torques_l2: -0.1495 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.1074 Episode_Reward/rolling_leg_motion: -0.7278 Episode_Reward/obstacle_progress: +2.8524 Episode_Reward/wheel_body_intrusion: -0.2859 Episode_Reward/pair_body_intrusion: -0.3564 Curriculum/terrain_level_mean: +2.8418 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0975 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.5069 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.4583 Episode_Termination/illegal_nonwheel_contact: 0.0139 ============================================================================== ============================================================================== Learning iteration 16900/50000 Total steps: 69,222,400 Steps per second: 11,521 Iteration time: 0.356s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0855 Mean episode length: 1326.0 Actor loss: -9.4868 Critic loss: +0.0042 World model loss: +0.8123 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5344 Episode_Reward/track_angular_velocity: +50.0692 Episode_Reward/upright: +25.5124 Episode_Reward/leg_posture: -2.7679 Episode_Reward/standstill: -6.5294 Episode_Reward/action_rate_l2: -0.6072 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.1102 Episode_Reward/rolling_leg_motion: -0.7173 Episode_Reward/obstacle_progress: +2.7738 Episode_Reward/wheel_body_intrusion: -0.2586 Episode_Reward/pair_body_intrusion: -0.3431 Curriculum/terrain_level_mean: +2.8484 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0987 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.4516 Episode_Termination/fell_over: 0.0323 Episode_Termination/course_exit: 0.4839 Episode_Termination/illegal_nonwheel_contact: 0.0387 ============================================================================== ============================================================================== Learning iteration 16950/50000 Total steps: 69,427,200 Steps per second: 11,530 Iteration time: 0.355s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0858 Mean episode length: 1381.9 Actor loss: -9.4552 Critic loss: +0.0055 World model loss: +0.7841 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5754 Episode_Reward/track_angular_velocity: +52.1638 Episode_Reward/upright: +26.6092 Episode_Reward/leg_posture: -2.8722 Episode_Reward/standstill: -6.5494 Episode_Reward/action_rate_l2: -0.6285 Episode_Reward/joint_torques_l2: -0.1508 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0243 Episode_Reward/unnecessary_wheel_lift: -0.1183 Episode_Reward/rolling_leg_motion: -0.7598 Episode_Reward/obstacle_progress: +2.8833 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.3585 Curriculum/terrain_level_mean: +2.8545 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0974 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.5159 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4268 Episode_Termination/illegal_nonwheel_contact: 0.0510 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 17000/50000 Total steps: 69,632,000 Steps per second: 11,538 Iteration time: 0.355s Collection time: 0.075s Learning time: 0.193s Mean reward: +0.0859 Mean episode length: 1364.8 Actor loss: -9.4805 Critic loss: +0.0100 World model loss: +0.8144 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1072 Episode_Reward/track_angular_velocity: +51.5253 Episode_Reward/upright: +26.2744 Episode_Reward/leg_posture: -2.8146 Episode_Reward/standstill: -6.5673 Episode_Reward/action_rate_l2: -0.6057 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.1015 Episode_Reward/rolling_leg_motion: -0.7296 Episode_Reward/obstacle_progress: +2.9694 Episode_Reward/wheel_body_intrusion: -0.2676 Episode_Reward/pair_body_intrusion: -0.3364 Curriculum/terrain_level_mean: +2.8594 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.5191 Episode_Termination/fell_over: 0.0458 Episode_Termination/course_exit: 0.4198 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 17050/50000 Total steps: 69,836,800 Steps per second: 11,547 Iteration time: 0.355s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0855 Mean episode length: 1355.2 Actor loss: -9.4597 Critic loss: +0.0032 World model loss: +0.7887 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5758 Episode_Reward/track_angular_velocity: +51.1664 Episode_Reward/upright: +26.0763 Episode_Reward/leg_posture: -2.8731 Episode_Reward/standstill: -6.8001 Episode_Reward/action_rate_l2: -0.6015 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.1167 Episode_Reward/rolling_leg_motion: -0.7039 Episode_Reward/obstacle_progress: +2.9819 Episode_Reward/wheel_body_intrusion: -0.2841 Episode_Reward/pair_body_intrusion: -0.3422 Curriculum/terrain_level_mean: +2.8665 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0986 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.4406 Episode_Termination/fell_over: 0.0280 Episode_Termination/course_exit: 0.5105 Episode_Termination/illegal_nonwheel_contact: 0.0420 ============================================================================== ============================================================================== Learning iteration 17100/50000 Total steps: 70,041,600 Steps per second: 11,556 Iteration time: 0.354s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0858 Mean episode length: 1310.3 Actor loss: -9.4208 Critic loss: +0.0056 World model loss: +0.8144 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1330 Episode_Reward/track_angular_velocity: +49.4599 Episode_Reward/upright: +25.2146 Episode_Reward/leg_posture: -2.7717 Episode_Reward/standstill: -6.3071 Episode_Reward/action_rate_l2: -0.5879 Episode_Reward/joint_torques_l2: -0.1461 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0230 Episode_Reward/unnecessary_wheel_lift: -0.1251 Episode_Reward/rolling_leg_motion: -0.7074 Episode_Reward/obstacle_progress: +2.8538 Episode_Reward/wheel_body_intrusion: -0.2700 Episode_Reward/pair_body_intrusion: -0.3338 Curriculum/terrain_level_mean: +2.8711 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.4729 Episode_Termination/fell_over: 0.0388 Episode_Termination/course_exit: 0.4574 Episode_Termination/illegal_nonwheel_contact: 0.0465 ============================================================================== ============================================================================== Learning iteration 17150/50000 Total steps: 70,246,400 Steps per second: 11,565 Iteration time: 0.354s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1330.7 Actor loss: -9.4698 Critic loss: +0.0046 World model loss: +0.7523 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8686 Episode_Reward/track_angular_velocity: +50.2856 Episode_Reward/upright: +25.6169 Episode_Reward/leg_posture: -2.8077 Episode_Reward/standstill: -5.9452 Episode_Reward/action_rate_l2: -0.5815 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0286 Episode_Reward/unnecessary_wheel_lift: -0.1108 Episode_Reward/rolling_leg_motion: -0.7059 Episode_Reward/obstacle_progress: +2.8582 Episode_Reward/wheel_body_intrusion: -0.2558 Episode_Reward/pair_body_intrusion: -0.3125 Curriculum/terrain_level_mean: +2.8774 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4786 Episode_Termination/fell_over: 0.0286 Episode_Termination/course_exit: 0.4429 Episode_Termination/illegal_nonwheel_contact: 0.0643 ============================================================================== ============================================================================== Learning iteration 17200/50000 Total steps: 70,451,200 Steps per second: 11,574 Iteration time: 0.354s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1356.9 Actor loss: -9.4675 Critic loss: +0.0069 World model loss: +0.7920 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8062 Episode_Reward/track_angular_velocity: +51.2614 Episode_Reward/upright: +26.1048 Episode_Reward/leg_posture: -2.8971 Episode_Reward/standstill: -6.1138 Episode_Reward/action_rate_l2: -0.5907 Episode_Reward/joint_torques_l2: -0.1507 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.1107 Episode_Reward/rolling_leg_motion: -0.7436 Episode_Reward/obstacle_progress: +2.9431 Episode_Reward/wheel_body_intrusion: -0.2785 Episode_Reward/pair_body_intrusion: -0.3017 Curriculum/terrain_level_mean: +2.8838 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.4815 Episode_Termination/fell_over: 0.0222 Episode_Termination/course_exit: 0.4889 Episode_Termination/illegal_nonwheel_contact: 0.0148 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1141] ============================================================================== Learning iteration 17250/50000 Total steps: 70,656,000 Steps per second: 11,583 Iteration time: 0.354s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0859 Mean episode length: 1398.4 Actor loss: -9.4616 Critic loss: +0.0051 World model loss: +0.8192 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1443 Episode_Reward/track_angular_velocity: +52.7967 Episode_Reward/upright: +26.9058 Episode_Reward/leg_posture: -2.9116 Episode_Reward/standstill: -6.6761 Episode_Reward/action_rate_l2: -0.6110 Episode_Reward/joint_torques_l2: -0.1543 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0273 Episode_Reward/unnecessary_wheel_lift: -0.1299 Episode_Reward/rolling_leg_motion: -0.7441 Episode_Reward/obstacle_progress: +3.0466 Episode_Reward/wheel_body_intrusion: -0.2803 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +2.8892 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0972 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.5157 Episode_Termination/fell_over: 0.0314 Episode_Termination/course_exit: 0.4214 Episode_Termination/illegal_nonwheel_contact: 0.0377 ============================================================================== ============================================================================== Learning iteration 17300/50000 Total steps: 70,860,800 Steps per second: 11,591 Iteration time: 0.353s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0859 Mean episode length: 1399.4 Actor loss: -9.4689 Critic loss: +0.0041 World model loss: +0.7969 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1739 Episode_Reward/track_angular_velocity: +52.8111 Episode_Reward/upright: +26.9351 Episode_Reward/leg_posture: -2.9461 Episode_Reward/standstill: -6.5528 Episode_Reward/action_rate_l2: -0.6280 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0246 Episode_Reward/unnecessary_wheel_lift: -0.1358 Episode_Reward/rolling_leg_motion: -0.7727 Episode_Reward/obstacle_progress: +3.0635 Episode_Reward/wheel_body_intrusion: -0.2946 Episode_Reward/pair_body_intrusion: -0.3110 Curriculum/terrain_level_mean: +2.8940 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0976 Metrics/base_velocity/error_vel_yaw: 0.1379 Episode_Termination/time_out: 0.5253 Episode_Termination/fell_over: 0.0506 Episode_Termination/course_exit: 0.4177 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== ============================================================================== Learning iteration 17350/50000 Total steps: 71,065,600 Steps per second: 11,600 Iteration time: 0.353s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0863 Mean episode length: 1418.3 Actor loss: -9.4701 Critic loss: +0.0037 World model loss: +0.7693 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8687 Episode_Reward/track_angular_velocity: +53.5512 Episode_Reward/upright: +27.3330 Episode_Reward/leg_posture: -2.9919 Episode_Reward/standstill: -6.0809 Episode_Reward/action_rate_l2: -0.6322 Episode_Reward/joint_torques_l2: -0.1572 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0194 Episode_Reward/unnecessary_wheel_lift: -0.1263 Episode_Reward/rolling_leg_motion: -0.7618 Episode_Reward/obstacle_progress: +3.0274 Episode_Reward/wheel_body_intrusion: -0.3079 Episode_Reward/pair_body_intrusion: -0.3026 Curriculum/terrain_level_mean: +2.9009 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0968 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.5280 Episode_Termination/fell_over: 0.0311 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0248 ============================================================================== ============================================================================== Learning iteration 17400/50000 Total steps: 71,270,400 Steps per second: 11,609 Iteration time: 0.353s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1359.2 Actor loss: -9.4782 Critic loss: +0.0037 World model loss: +0.7704 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9186 Episode_Reward/track_angular_velocity: +51.3672 Episode_Reward/upright: +26.2002 Episode_Reward/leg_posture: -2.8398 Episode_Reward/standstill: -6.1485 Episode_Reward/action_rate_l2: -0.6124 Episode_Reward/joint_torques_l2: -0.1502 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.1294 Episode_Reward/rolling_leg_motion: -0.7201 Episode_Reward/obstacle_progress: +2.9892 Episode_Reward/wheel_body_intrusion: -0.2627 Episode_Reward/pair_body_intrusion: -0.3184 Curriculum/terrain_level_mean: +2.9070 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0961 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4792 Episode_Termination/fell_over: 0.0139 Episode_Termination/course_exit: 0.4792 Episode_Termination/illegal_nonwheel_contact: 0.0347 ============================================================================== ============================================================================== Learning iteration 17450/50000 Total steps: 71,475,200 Steps per second: 11,618 Iteration time: 0.353s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1361.4 Actor loss: -9.4844 Critic loss: +0.0076 World model loss: +0.7757 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8957 Episode_Reward/track_angular_velocity: +51.4178 Episode_Reward/upright: +26.2173 Episode_Reward/leg_posture: -2.8688 Episode_Reward/standstill: -6.2208 Episode_Reward/action_rate_l2: -0.6029 Episode_Reward/joint_torques_l2: -0.1508 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.1279 Episode_Reward/rolling_leg_motion: -0.7306 Episode_Reward/obstacle_progress: +2.8934 Episode_Reward/wheel_body_intrusion: -0.2683 Episode_Reward/pair_body_intrusion: -0.3212 Curriculum/terrain_level_mean: +2.9106 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0968 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.5338 Episode_Termination/fell_over: 0.0135 Episode_Termination/course_exit: 0.4324 Episode_Termination/illegal_nonwheel_contact: 0.0270 ============================================================================== ============================================================================== Learning iteration 17500/50000 Total steps: 71,680,000 Steps per second: 11,627 Iteration time: 0.352s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1420.3 Actor loss: -9.4437 Critic loss: +0.0042 World model loss: +0.7972 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0843 Episode_Reward/track_angular_velocity: +53.6865 Episode_Reward/upright: +27.3587 Episode_Reward/leg_posture: -2.9885 Episode_Reward/standstill: -6.3785 Episode_Reward/action_rate_l2: -0.6356 Episode_Reward/joint_torques_l2: -0.1565 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0222 Episode_Reward/unnecessary_wheel_lift: -0.1138 Episode_Reward/rolling_leg_motion: -0.7934 Episode_Reward/obstacle_progress: +3.0157 Episode_Reward/wheel_body_intrusion: -0.2927 Episode_Reward/pair_body_intrusion: -0.3220 Curriculum/terrain_level_mean: +2.9189 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5399 Episode_Termination/fell_over: 0.0245 Episode_Termination/course_exit: 0.4110 Episode_Termination/illegal_nonwheel_contact: 0.0307 ============================================================================== ============================================================================== Learning iteration 17550/50000 Total steps: 71,884,800 Steps per second: 11,635 Iteration time: 0.352s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0863 Mean episode length: 1381.6 Actor loss: -9.4562 Critic loss: +0.0055 World model loss: +0.7917 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8625 Episode_Reward/track_angular_velocity: +52.2569 Episode_Reward/upright: +26.6282 Episode_Reward/leg_posture: -2.8788 Episode_Reward/standstill: -6.1782 Episode_Reward/action_rate_l2: -0.6061 Episode_Reward/joint_torques_l2: -0.1534 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.1343 Episode_Reward/rolling_leg_motion: -0.7679 Episode_Reward/obstacle_progress: +2.8584 Episode_Reward/wheel_body_intrusion: -0.2892 Episode_Reward/pair_body_intrusion: -0.3297 Curriculum/terrain_level_mean: +2.9255 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4771 Episode_Termination/fell_over: 0.0196 Episode_Termination/course_exit: 0.4837 Episode_Termination/illegal_nonwheel_contact: 0.0261 ============================================================================== ============================================================================== Learning iteration 17600/50000 Total steps: 72,089,600 Steps per second: 11,643 Iteration time: 0.352s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1351.3 Actor loss: -9.4852 Critic loss: +0.0036 World model loss: +0.7882 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7799 Episode_Reward/track_angular_velocity: +51.1280 Episode_Reward/upright: +26.0361 Episode_Reward/leg_posture: -2.7836 Episode_Reward/standstill: -6.3521 Episode_Reward/action_rate_l2: -0.5902 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.1105 Episode_Reward/rolling_leg_motion: -0.7301 Episode_Reward/obstacle_progress: +2.7844 Episode_Reward/wheel_body_intrusion: -0.2642 Episode_Reward/pair_body_intrusion: -0.3060 Curriculum/terrain_level_mean: +2.9324 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4733 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.4933 Episode_Termination/illegal_nonwheel_contact: 0.0200 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[960] ============================================================================== Learning iteration 17650/50000 Total steps: 72,294,400 Steps per second: 11,652 Iteration time: 0.352s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1322.7 Actor loss: -9.4994 Critic loss: +0.0043 World model loss: +0.7719 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7435 Episode_Reward/track_angular_velocity: +50.0004 Episode_Reward/upright: +25.4641 Episode_Reward/leg_posture: -2.7168 Episode_Reward/standstill: -6.4028 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1463 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.1115 Episode_Reward/rolling_leg_motion: -0.7202 Episode_Reward/obstacle_progress: +2.8287 Episode_Reward/wheel_body_intrusion: -0.2536 Episode_Reward/pair_body_intrusion: -0.3207 Curriculum/terrain_level_mean: +2.9382 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4780 Episode_Termination/fell_over: 0.0629 Episode_Termination/course_exit: 0.4528 Episode_Termination/illegal_nonwheel_contact: 0.0377 ============================================================================== ============================================================================== Learning iteration 17700/50000 Total steps: 72,499,200 Steps per second: 11,660 Iteration time: 0.351s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1314.7 Actor loss: -9.4606 Critic loss: +0.0081 World model loss: +0.7670 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3226 Episode_Reward/track_angular_velocity: +49.6954 Episode_Reward/upright: +25.2788 Episode_Reward/leg_posture: -2.7472 Episode_Reward/standstill: -6.1342 Episode_Reward/action_rate_l2: -0.5736 Episode_Reward/joint_torques_l2: -0.1452 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0128 Episode_Reward/unnecessary_wheel_lift: -0.1159 Episode_Reward/rolling_leg_motion: -0.7217 Episode_Reward/obstacle_progress: +2.7417 Episode_Reward/wheel_body_intrusion: -0.2683 Episode_Reward/pair_body_intrusion: -0.3203 Curriculum/terrain_level_mean: +2.9465 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0961 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0179 Episode_Termination/course_exit: 0.5417 Episode_Termination/illegal_nonwheel_contact: 0.0179 ============================================================================== ============================================================================== Learning iteration 17750/50000 Total steps: 72,704,000 Steps per second: 11,669 Iteration time: 0.351s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0859 Mean episode length: 1371.8 Actor loss: -9.4491 Critic loss: +0.0068 World model loss: +0.7816 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2380 Episode_Reward/track_angular_velocity: +51.8798 Episode_Reward/upright: +26.4125 Episode_Reward/leg_posture: -2.8271 Episode_Reward/standstill: -6.3365 Episode_Reward/action_rate_l2: -0.6257 Episode_Reward/joint_torques_l2: -0.1518 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.1336 Episode_Reward/rolling_leg_motion: -0.7597 Episode_Reward/obstacle_progress: +2.7787 Episode_Reward/wheel_body_intrusion: -0.2875 Episode_Reward/pair_body_intrusion: -0.3322 Curriculum/terrain_level_mean: +2.9519 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4934 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.4605 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== ============================================================================== Learning iteration 17800/50000 Total steps: 72,908,800 Steps per second: 11,677 Iteration time: 0.351s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1354.1 Actor loss: -9.4816 Critic loss: +0.0068 World model loss: +0.7798 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8089 Episode_Reward/track_angular_velocity: +51.1693 Episode_Reward/upright: +26.0856 Episode_Reward/leg_posture: -2.7755 Episode_Reward/standstill: -6.3229 Episode_Reward/action_rate_l2: -0.6016 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0053 Episode_Reward/unnecessary_wheel_lift: -0.0936 Episode_Reward/rolling_leg_motion: -0.7232 Episode_Reward/obstacle_progress: +2.8142 Episode_Reward/wheel_body_intrusion: -0.2802 Episode_Reward/pair_body_intrusion: -0.3573 Curriculum/terrain_level_mean: +2.9590 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0949 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.5156 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4688 Episode_Termination/illegal_nonwheel_contact: 0.0156 ============================================================================== ============================================================================== Learning iteration 17850/50000 Total steps: 73,113,600 Steps per second: 11,686 Iteration time: 0.351s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0858 Mean episode length: 1359.6 Actor loss: -9.4831 Critic loss: +0.0039 World model loss: +0.7714 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7527 Episode_Reward/track_angular_velocity: +51.3566 Episode_Reward/upright: +26.1629 Episode_Reward/leg_posture: -2.8118 Episode_Reward/standstill: -6.3657 Episode_Reward/action_rate_l2: -0.6103 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0186 Episode_Reward/unnecessary_wheel_lift: -0.1157 Episode_Reward/rolling_leg_motion: -0.7320 Episode_Reward/obstacle_progress: +2.8275 Episode_Reward/wheel_body_intrusion: -0.2779 Episode_Reward/pair_body_intrusion: -0.3226 Curriculum/terrain_level_mean: +2.9641 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0974 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.5355 Episode_Termination/fell_over: 0.0065 Episode_Termination/course_exit: 0.4194 Episode_Termination/illegal_nonwheel_contact: 0.0452 ============================================================================== ============================================================================== Learning iteration 17900/50000 Total steps: 73,318,400 Steps per second: 11,694 Iteration time: 0.350s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0857 Mean episode length: 1388.8 Actor loss: -9.4715 Critic loss: +0.0038 World model loss: +0.8082 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8258 Episode_Reward/track_angular_velocity: +52.4859 Episode_Reward/upright: +26.7071 Episode_Reward/leg_posture: -2.9338 Episode_Reward/standstill: -6.5882 Episode_Reward/action_rate_l2: -0.6179 Episode_Reward/joint_torques_l2: -0.1543 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.1259 Episode_Reward/rolling_leg_motion: -0.7425 Episode_Reward/obstacle_progress: +2.7781 Episode_Reward/wheel_body_intrusion: -0.2882 Episode_Reward/pair_body_intrusion: -0.3356 Curriculum/terrain_level_mean: +2.9692 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4819 Episode_Termination/fell_over: 0.0422 Episode_Termination/course_exit: 0.4578 Episode_Termination/illegal_nonwheel_contact: 0.0361 ============================================================================== ============================================================================== Learning iteration 17950/50000 Total steps: 73,523,200 Steps per second: 11,703 Iteration time: 0.350s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1399.2 Actor loss: -9.4560 Critic loss: +0.0029 World model loss: +0.7329 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2238 Episode_Reward/track_angular_velocity: +52.9420 Episode_Reward/upright: +26.8982 Episode_Reward/leg_posture: -2.9586 Episode_Reward/standstill: -6.3082 Episode_Reward/action_rate_l2: -0.6122 Episode_Reward/joint_torques_l2: -0.1554 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0273 Episode_Reward/unnecessary_wheel_lift: -0.1177 Episode_Reward/rolling_leg_motion: -0.7506 Episode_Reward/obstacle_progress: +2.8557 Episode_Reward/wheel_body_intrusion: -0.2886 Episode_Reward/pair_body_intrusion: -0.3212 Curriculum/terrain_level_mean: +2.9751 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5200 Episode_Termination/fell_over: 0.0400 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== ============================================================================== Learning iteration 18000/50000 Total steps: 73,728,000 Steps per second: 11,711 Iteration time: 0.350s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0857 Mean episode length: 1410.2 Actor loss: -9.4652 Critic loss: +0.0032 World model loss: +0.7971 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5734 Episode_Reward/track_angular_velocity: +53.2455 Episode_Reward/upright: +27.1270 Episode_Reward/leg_posture: -3.0165 Episode_Reward/standstill: -6.4887 Episode_Reward/action_rate_l2: -0.6060 Episode_Reward/joint_torques_l2: -0.1572 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0193 Episode_Reward/unnecessary_wheel_lift: -0.1218 Episode_Reward/rolling_leg_motion: -0.7754 Episode_Reward/obstacle_progress: +2.8302 Episode_Reward/wheel_body_intrusion: -0.2972 Episode_Reward/pair_body_intrusion: -0.3692 Curriculum/terrain_level_mean: +2.9812 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0970 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.5333 Episode_Termination/fell_over: 0.0444 Episode_Termination/course_exit: 0.4148 Episode_Termination/illegal_nonwheel_contact: 0.0148 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1104] ============================================================================== Learning iteration 18050/50000 Total steps: 73,932,800 Steps per second: 11,719 Iteration time: 0.350s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0858 Mean episode length: 1346.6 Actor loss: -9.4590 Critic loss: +0.0031 World model loss: +0.7711 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4100 Episode_Reward/track_angular_velocity: +50.8328 Episode_Reward/upright: +25.9309 Episode_Reward/leg_posture: -2.8524 Episode_Reward/standstill: -6.3736 Episode_Reward/action_rate_l2: -0.5929 Episode_Reward/joint_torques_l2: -0.1494 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1095 Episode_Reward/rolling_leg_motion: -0.7331 Episode_Reward/obstacle_progress: +2.7605 Episode_Reward/wheel_body_intrusion: -0.2787 Episode_Reward/pair_body_intrusion: -0.2929 Curriculum/terrain_level_mean: +2.9895 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4967 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.4636 Episode_Termination/illegal_nonwheel_contact: 0.0331 ============================================================================== ============================================================================== Learning iteration 18100/50000 Total steps: 74,137,600 Steps per second: 11,727 Iteration time: 0.349s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0858 Mean episode length: 1351.4 Actor loss: -9.4652 Critic loss: +0.0061 World model loss: +0.7483 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4546 Episode_Reward/track_angular_velocity: +50.9856 Episode_Reward/upright: +25.9943 Episode_Reward/leg_posture: -2.8414 Episode_Reward/standstill: -6.1923 Episode_Reward/action_rate_l2: -0.5832 Episode_Reward/joint_torques_l2: -0.1518 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0290 Episode_Reward/unnecessary_wheel_lift: -0.1226 Episode_Reward/rolling_leg_motion: -0.7162 Episode_Reward/obstacle_progress: +2.7165 Episode_Reward/wheel_body_intrusion: -0.2684 Episode_Reward/pair_body_intrusion: -0.2870 Curriculum/terrain_level_mean: +2.9971 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0976 Metrics/base_velocity/error_vel_yaw: 0.1378 Episode_Termination/time_out: 0.4753 Episode_Termination/fell_over: 0.0494 Episode_Termination/course_exit: 0.4568 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2126] RECOVERY reset 1 invalid environments; first IDs=[628] ============================================================================== Learning iteration 18150/50000 Total steps: 74,342,400 Steps per second: 11,735 Iteration time: 0.349s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0863 Mean episode length: 1385.4 Actor loss: -9.4660 Critic loss: +0.0044 World model loss: +0.8018 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8834 Episode_Reward/track_angular_velocity: +52.2907 Episode_Reward/upright: +26.6827 Episode_Reward/leg_posture: -2.9277 Episode_Reward/standstill: -6.1016 Episode_Reward/action_rate_l2: -0.5951 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1119 Episode_Reward/rolling_leg_motion: -0.7429 Episode_Reward/obstacle_progress: +2.9114 Episode_Reward/wheel_body_intrusion: -0.2787 Episode_Reward/pair_body_intrusion: -0.3273 Curriculum/terrain_level_mean: +3.0010 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0954 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.5147 Episode_Termination/fell_over: 0.0441 Episode_Termination/course_exit: 0.4338 Episode_Termination/illegal_nonwheel_contact: 0.0221 ============================================================================== ============================================================================== Learning iteration 18200/50000 Total steps: 74,547,200 Steps per second: 11,744 Iteration time: 0.349s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1365.7 Actor loss: -9.4663 Critic loss: +0.0034 World model loss: +0.7682 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1116 Episode_Reward/track_angular_velocity: +51.5441 Episode_Reward/upright: +26.3025 Episode_Reward/leg_posture: -2.8267 Episode_Reward/standstill: -6.2806 Episode_Reward/action_rate_l2: -0.5837 Episode_Reward/joint_torques_l2: -0.1499 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0981 Episode_Reward/rolling_leg_motion: -0.7283 Episode_Reward/obstacle_progress: +2.8431 Episode_Reward/wheel_body_intrusion: -0.2707 Episode_Reward/pair_body_intrusion: -0.3871 Curriculum/terrain_level_mean: +3.0090 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0959 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.4830 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4966 Episode_Termination/illegal_nonwheel_contact: 0.0204 ============================================================================== ============================================================================== Learning iteration 18250/50000 Total steps: 74,752,000 Steps per second: 11,752 Iteration time: 0.349s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0860 Mean episode length: 1370.2 Actor loss: -9.4866 Critic loss: +0.0027 World model loss: +0.7602 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3400 Episode_Reward/track_angular_velocity: +51.7568 Episode_Reward/upright: +26.3787 Episode_Reward/leg_posture: -2.8987 Episode_Reward/standstill: -6.3931 Episode_Reward/action_rate_l2: -0.5900 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1171 Episode_Reward/rolling_leg_motion: -0.7426 Episode_Reward/obstacle_progress: +2.8737 Episode_Reward/wheel_body_intrusion: -0.2763 Episode_Reward/pair_body_intrusion: -0.3400 Curriculum/terrain_level_mean: +3.0132 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0953 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4792 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.4653 Episode_Termination/illegal_nonwheel_contact: 0.0417 ============================================================================== ============================================================================== Learning iteration 18300/50000 Total steps: 74,956,800 Steps per second: 11,760 Iteration time: 0.348s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0866 Mean episode length: 1339.2 Actor loss: -9.4613 Critic loss: +0.0035 World model loss: +0.7660 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3421 Episode_Reward/track_angular_velocity: +50.6702 Episode_Reward/upright: +25.8085 Episode_Reward/leg_posture: -2.7938 Episode_Reward/standstill: -5.8303 Episode_Reward/action_rate_l2: -0.5739 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.1012 Episode_Reward/rolling_leg_motion: -0.6952 Episode_Reward/obstacle_progress: +2.8763 Episode_Reward/wheel_body_intrusion: -0.2639 Episode_Reward/pair_body_intrusion: -0.3082 Curriculum/terrain_level_mean: +3.0210 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4276 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.5263 Episode_Termination/illegal_nonwheel_contact: 0.0329 ============================================================================== ============================================================================== Learning iteration 18350/50000 Total steps: 75,161,600 Steps per second: 11,767 Iteration time: 0.348s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0864 Mean episode length: 1424.6 Actor loss: -9.4591 Critic loss: +0.0063 World model loss: +0.7407 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2846 Episode_Reward/track_angular_velocity: +53.8451 Episode_Reward/upright: +27.4331 Episode_Reward/leg_posture: -2.9975 Episode_Reward/standstill: -6.3570 Episode_Reward/action_rate_l2: -0.6104 Episode_Reward/joint_torques_l2: -0.1560 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0250 Episode_Reward/unnecessary_wheel_lift: -0.1182 Episode_Reward/rolling_leg_motion: -0.7432 Episode_Reward/obstacle_progress: +3.0752 Episode_Reward/wheel_body_intrusion: -0.2917 Episode_Reward/pair_body_intrusion: -0.2986 Curriculum/terrain_level_mean: +3.0239 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0951 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5570 Episode_Termination/fell_over: 0.0316 Episode_Termination/course_exit: 0.3861 Episode_Termination/illegal_nonwheel_contact: 0.0316 ============================================================================== ============================================================================== Learning iteration 18400/50000 Total steps: 75,366,400 Steps per second: 11,775 Iteration time: 0.348s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0865 Mean episode length: 1404.4 Actor loss: -9.4435 Critic loss: +0.0067 World model loss: +0.7801 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5359 Episode_Reward/track_angular_velocity: +53.1004 Episode_Reward/upright: +27.0553 Episode_Reward/leg_posture: -2.9110 Episode_Reward/standstill: -6.1970 Episode_Reward/action_rate_l2: -0.6157 Episode_Reward/joint_torques_l2: -0.1531 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.1093 Episode_Reward/rolling_leg_motion: -0.7359 Episode_Reward/obstacle_progress: +3.0534 Episode_Reward/wheel_body_intrusion: -0.2739 Episode_Reward/pair_body_intrusion: -0.3204 Curriculum/terrain_level_mean: +3.0283 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5139 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.4514 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== ============================================================================== Learning iteration 18450/50000 Total steps: 75,571,200 Steps per second: 11,783 Iteration time: 0.348s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0856 Mean episode length: 1372.8 Actor loss: -9.4585 Critic loss: +0.0103 World model loss: +0.7875 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2482 Episode_Reward/track_angular_velocity: +51.8216 Episode_Reward/upright: +26.4167 Episode_Reward/leg_posture: -2.8578 Episode_Reward/standstill: -6.8847 Episode_Reward/action_rate_l2: -0.6196 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1240 Episode_Reward/rolling_leg_motion: -0.7205 Episode_Reward/obstacle_progress: +2.9611 Episode_Reward/wheel_body_intrusion: -0.2700 Episode_Reward/pair_body_intrusion: -0.3177 Curriculum/terrain_level_mean: +3.0317 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0969 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.5076 Episode_Termination/fell_over: 0.0379 Episode_Termination/course_exit: 0.4318 Episode_Termination/illegal_nonwheel_contact: 0.0379 ============================================================================== ============================================================================== Learning iteration 18500/50000 Total steps: 75,776,000 Steps per second: 11,792 Iteration time: 0.347s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0863 Mean episode length: 1353.5 Actor loss: -9.4551 Critic loss: +0.0054 World model loss: +0.7675 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7356 Episode_Reward/track_angular_velocity: +51.1895 Episode_Reward/upright: +26.0638 Episode_Reward/leg_posture: -2.7914 Episode_Reward/standstill: -6.1954 Episode_Reward/action_rate_l2: -0.5897 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0264 Episode_Reward/unnecessary_wheel_lift: -0.1083 Episode_Reward/rolling_leg_motion: -0.7079 Episode_Reward/obstacle_progress: +2.9665 Episode_Reward/wheel_body_intrusion: -0.2575 Episode_Reward/pair_body_intrusion: -0.3123 Curriculum/terrain_level_mean: +3.0383 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0953 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4884 Episode_Termination/fell_over: 0.0291 Episode_Termination/course_exit: 0.4535 Episode_Termination/illegal_nonwheel_contact: 0.0407 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1762] ============================================================================== Learning iteration 18550/50000 Total steps: 75,980,800 Steps per second: 11,800 Iteration time: 0.347s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0859 Mean episode length: 1308.2 Actor loss: -9.4818 Critic loss: +0.0055 World model loss: +0.7237 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0805 Episode_Reward/track_angular_velocity: +49.3733 Episode_Reward/upright: +25.1776 Episode_Reward/leg_posture: -2.7114 Episode_Reward/standstill: -6.3041 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1438 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0256 Episode_Reward/unnecessary_wheel_lift: -0.1013 Episode_Reward/rolling_leg_motion: -0.6918 Episode_Reward/obstacle_progress: +2.9133 Episode_Reward/wheel_body_intrusion: -0.2657 Episode_Reward/pair_body_intrusion: -0.3142 Curriculum/terrain_level_mean: +3.0439 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0544 ============================================================================== ============================================================================== Learning iteration 18600/50000 Total steps: 76,185,600 Steps per second: 11,807 Iteration time: 0.347s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1322.4 Actor loss: -9.4981 Critic loss: +0.0028 World model loss: +0.7989 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5626 Episode_Reward/track_angular_velocity: +49.9120 Episode_Reward/upright: +25.4665 Episode_Reward/leg_posture: -2.7440 Episode_Reward/standstill: -6.1403 Episode_Reward/action_rate_l2: -0.5716 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0927 Episode_Reward/rolling_leg_motion: -0.6745 Episode_Reward/obstacle_progress: +3.0026 Episode_Reward/wheel_body_intrusion: -0.2766 Episode_Reward/pair_body_intrusion: -0.2984 Curriculum/terrain_level_mean: +3.0505 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0203 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== ============================================================================== Learning iteration 18650/50000 Total steps: 76,390,400 Steps per second: 11,815 Iteration time: 0.347s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1350.4 Actor loss: -9.4765 Critic loss: +0.0033 World model loss: +0.7598 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5352 Episode_Reward/track_angular_velocity: +50.9775 Episode_Reward/upright: +25.9815 Episode_Reward/leg_posture: -2.7711 Episode_Reward/standstill: -6.3417 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0264 Episode_Reward/unnecessary_wheel_lift: -0.1004 Episode_Reward/rolling_leg_motion: -0.7003 Episode_Reward/obstacle_progress: +3.1017 Episode_Reward/wheel_body_intrusion: -0.2635 Episode_Reward/pair_body_intrusion: -0.3329 Curriculum/terrain_level_mean: +3.0535 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0961 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.5182 Episode_Termination/fell_over: 0.0219 Episode_Termination/course_exit: 0.4088 Episode_Termination/illegal_nonwheel_contact: 0.0584 ============================================================================== ============================================================================== Learning iteration 18700/50000 Total steps: 76,595,200 Steps per second: 11,823 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0866 Mean episode length: 1335.4 Actor loss: -9.4576 Critic loss: +0.0036 World model loss: +0.7756 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2274 Episode_Reward/track_angular_velocity: +50.5107 Episode_Reward/upright: +25.7206 Episode_Reward/leg_posture: -2.7299 Episode_Reward/standstill: -6.0588 Episode_Reward/action_rate_l2: -0.5837 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0287 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.6690 Episode_Reward/obstacle_progress: +3.0224 Episode_Reward/wheel_body_intrusion: -0.2727 Episode_Reward/pair_body_intrusion: -0.2927 Curriculum/terrain_level_mean: +3.0566 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.5231 Episode_Termination/fell_over: 0.0231 Episode_Termination/course_exit: 0.3923 Episode_Termination/illegal_nonwheel_contact: 0.0769 ============================================================================== ============================================================================== Learning iteration 18750/50000 Total steps: 76,800,000 Steps per second: 11,831 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1330.7 Actor loss: -9.4617 Critic loss: +0.0053 World model loss: +0.7218 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9508 Episode_Reward/track_angular_velocity: +50.3080 Episode_Reward/upright: +25.6153 Episode_Reward/leg_posture: -2.7439 Episode_Reward/standstill: -6.1328 Episode_Reward/action_rate_l2: -0.5718 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.1044 Episode_Reward/rolling_leg_motion: -0.6932 Episode_Reward/obstacle_progress: +2.8576 Episode_Reward/wheel_body_intrusion: -0.2593 Episode_Reward/pair_body_intrusion: -0.3325 Curriculum/terrain_level_mean: +3.0662 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0949 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4189 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.5405 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== ============================================================================== Learning iteration 18800/50000 Total steps: 77,004,800 Steps per second: 11,839 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1296.3 Actor loss: -9.4568 Critic loss: +0.0054 World model loss: +0.7470 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7821 Episode_Reward/track_angular_velocity: +48.9939 Episode_Reward/upright: +24.9626 Episode_Reward/leg_posture: -2.6660 Episode_Reward/standstill: -6.2707 Episode_Reward/action_rate_l2: -0.5765 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0329 Episode_Reward/unnecessary_wheel_lift: -0.1182 Episode_Reward/rolling_leg_motion: -0.6970 Episode_Reward/obstacle_progress: +2.8099 Episode_Reward/wheel_body_intrusion: -0.2439 Episode_Reward/pair_body_intrusion: -0.3103 Curriculum/terrain_level_mean: +3.0720 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4277 Episode_Termination/fell_over: 0.0361 Episode_Termination/course_exit: 0.4940 Episode_Termination/illegal_nonwheel_contact: 0.0602 ============================================================================== ============================================================================== Learning iteration 18850/50000 Total steps: 77,209,600 Steps per second: 11,846 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0859 Mean episode length: 1309.4 Actor loss: -9.4433 Critic loss: +0.0064 World model loss: +0.7632 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3266 Episode_Reward/track_angular_velocity: +49.4907 Episode_Reward/upright: +25.2175 Episode_Reward/leg_posture: -2.7109 Episode_Reward/standstill: -6.5816 Episode_Reward/action_rate_l2: -0.5613 Episode_Reward/joint_torques_l2: -0.1436 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0281 Episode_Reward/unnecessary_wheel_lift: -0.1024 Episode_Reward/rolling_leg_motion: -0.6793 Episode_Reward/obstacle_progress: +2.8262 Episode_Reward/wheel_body_intrusion: -0.2711 Episode_Reward/pair_body_intrusion: -0.3494 Curriculum/terrain_level_mean: +3.0747 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4365 Episode_Termination/fell_over: 0.0556 Episode_Termination/course_exit: 0.4762 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2376] ============================================================================== Learning iteration 18900/50000 Total steps: 77,414,400 Steps per second: 11,854 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0859 Mean episode length: 1337.6 Actor loss: -9.5165 Critic loss: +0.0027 World model loss: +0.7608 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0849 Episode_Reward/track_angular_velocity: +50.4819 Episode_Reward/upright: +25.7629 Episode_Reward/leg_posture: -2.7601 Episode_Reward/standstill: -6.4713 Episode_Reward/action_rate_l2: -0.5786 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.0961 Episode_Reward/rolling_leg_motion: -0.6978 Episode_Reward/obstacle_progress: +2.9340 Episode_Reward/wheel_body_intrusion: -0.2762 Episode_Reward/pair_body_intrusion: -0.3691 Curriculum/terrain_level_mean: +3.0811 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4626 Episode_Termination/fell_over: 0.0136 Episode_Termination/course_exit: 0.4898 Episode_Termination/illegal_nonwheel_contact: 0.0340 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_018900.mp4 ============================================================================== Learning iteration 18950/50000 Total steps: 77,619,200 Steps per second: 11,851 Iteration time: 0.346s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0854 Mean episode length: 1305.1 Actor loss: -9.4691 Critic loss: +0.0106 World model loss: +0.7348 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8292 Episode_Reward/track_angular_velocity: +49.2211 Episode_Reward/upright: +25.0869 Episode_Reward/leg_posture: -2.7268 Episode_Reward/standstill: -6.6584 Episode_Reward/action_rate_l2: -0.5738 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0331 Episode_Reward/unnecessary_wheel_lift: -0.1064 Episode_Reward/rolling_leg_motion: -0.6858 Episode_Reward/obstacle_progress: +2.8429 Episode_Reward/wheel_body_intrusion: -0.2605 Episode_Reward/pair_body_intrusion: -0.3454 Curriculum/terrain_level_mean: +3.0854 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0973 Metrics/base_velocity/error_vel_yaw: 0.1378 Episode_Termination/time_out: 0.4658 Episode_Termination/fell_over: 0.0616 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 19000/50000 Total steps: 77,824,000 Steps per second: 11,858 Iteration time: 0.345s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0859 Mean episode length: 1395.5 Actor loss: -9.4648 Critic loss: +0.0034 World model loss: +0.7317 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1725 Episode_Reward/track_angular_velocity: +52.6483 Episode_Reward/upright: +26.8208 Episode_Reward/leg_posture: -2.9370 Episode_Reward/standstill: -6.5247 Episode_Reward/action_rate_l2: -0.6111 Episode_Reward/joint_torques_l2: -0.1541 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0327 Episode_Reward/unnecessary_wheel_lift: -0.1191 Episode_Reward/rolling_leg_motion: -0.7320 Episode_Reward/obstacle_progress: +3.0163 Episode_Reward/wheel_body_intrusion: -0.2656 Episode_Reward/pair_body_intrusion: -0.3799 Curriculum/terrain_level_mean: +3.0874 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0959 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.4720 Episode_Termination/fell_over: 0.0559 Episode_Termination/course_exit: 0.4534 Episode_Termination/illegal_nonwheel_contact: 0.0559 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1857] ============================================================================== Learning iteration 19050/50000 Total steps: 78,028,800 Steps per second: 11,866 Iteration time: 0.345s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0861 Mean episode length: 1366.9 Actor loss: -9.4645 Critic loss: +0.0031 World model loss: +0.8035 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2707 Episode_Reward/track_angular_velocity: +51.6583 Episode_Reward/upright: +26.3360 Episode_Reward/leg_posture: -2.8512 Episode_Reward/standstill: -6.5040 Episode_Reward/action_rate_l2: -0.5736 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1131 Episode_Reward/rolling_leg_motion: -0.7064 Episode_Reward/obstacle_progress: +2.9701 Episode_Reward/wheel_body_intrusion: -0.2766 Episode_Reward/pair_body_intrusion: -0.3534 Curriculum/terrain_level_mean: +3.0916 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5232 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.4238 Episode_Termination/illegal_nonwheel_contact: 0.0530 ============================================================================== ============================================================================== Learning iteration 19100/50000 Total steps: 78,233,600 Steps per second: 11,873 Iteration time: 0.345s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1346.2 Actor loss: -9.4705 Critic loss: +0.0037 World model loss: +0.7566 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5838 Episode_Reward/track_angular_velocity: +50.9201 Episode_Reward/upright: +25.9327 Episode_Reward/leg_posture: -2.7912 Episode_Reward/standstill: -6.2974 Episode_Reward/action_rate_l2: -0.5866 Episode_Reward/joint_torques_l2: -0.1502 Episode_Reward/dof_pos_limits: -0.0001 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1335 Episode_Reward/rolling_leg_motion: -0.7218 Episode_Reward/obstacle_progress: +2.8801 Episode_Reward/wheel_body_intrusion: -0.2595 Episode_Reward/pair_body_intrusion: -0.3270 Curriculum/terrain_level_mean: +3.0979 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5301 Episode_Termination/fell_over: 0.0181 Episode_Termination/course_exit: 0.4337 Episode_Termination/illegal_nonwheel_contact: 0.0241 ============================================================================== ============================================================================== Learning iteration 19150/50000 Total steps: 78,438,400 Steps per second: 11,881 Iteration time: 0.345s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1409.2 Actor loss: -9.4352 Critic loss: +0.0057 World model loss: +0.7589 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8190 Episode_Reward/track_angular_velocity: +53.3059 Episode_Reward/upright: +27.1282 Episode_Reward/leg_posture: -2.9086 Episode_Reward/standstill: -6.7651 Episode_Reward/action_rate_l2: -0.6070 Episode_Reward/joint_torques_l2: -0.1549 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.1331 Episode_Reward/rolling_leg_motion: -0.7620 Episode_Reward/obstacle_progress: +3.0866 Episode_Reward/wheel_body_intrusion: -0.2798 Episode_Reward/pair_body_intrusion: -0.3249 Curriculum/terrain_level_mean: +3.1042 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.5287 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.4331 Episode_Termination/illegal_nonwheel_contact: 0.0255 ============================================================================== ============================================================================== Learning iteration 19200/50000 Total steps: 78,643,200 Steps per second: 11,889 Iteration time: 0.345s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0858 Mean episode length: 1362.2 Actor loss: -9.4570 Critic loss: +0.0053 World model loss: +0.7824 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0274 Episode_Reward/track_angular_velocity: +51.4950 Episode_Reward/upright: +26.2014 Episode_Reward/leg_posture: -2.8586 Episode_Reward/standstill: -6.6648 Episode_Reward/action_rate_l2: -0.5929 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.1178 Episode_Reward/rolling_leg_motion: -0.7271 Episode_Reward/obstacle_progress: +2.9789 Episode_Reward/wheel_body_intrusion: -0.2842 Episode_Reward/pair_body_intrusion: -0.3928 Curriculum/terrain_level_mean: +3.1086 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0952 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4777 Episode_Termination/fell_over: 0.0382 Episode_Termination/course_exit: 0.4713 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== ============================================================================== Learning iteration 19250/50000 Total steps: 78,848,000 Steps per second: 11,896 Iteration time: 0.344s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1337.1 Actor loss: -9.4666 Critic loss: +0.0072 World model loss: +0.7589 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2991 Episode_Reward/track_angular_velocity: +50.5333 Episode_Reward/upright: +25.6944 Episode_Reward/leg_posture: -2.7656 Episode_Reward/standstill: -6.4847 Episode_Reward/action_rate_l2: -0.5822 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0287 Episode_Reward/unnecessary_wheel_lift: -0.1190 Episode_Reward/rolling_leg_motion: -0.7112 Episode_Reward/obstacle_progress: +2.9597 Episode_Reward/wheel_body_intrusion: -0.2456 Episode_Reward/pair_body_intrusion: -0.3361 Curriculum/terrain_level_mean: +3.1147 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4525 Episode_Termination/fell_over: 0.0223 Episode_Termination/course_exit: 0.4860 Episode_Termination/illegal_nonwheel_contact: 0.0503 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1372] ============================================================================== Learning iteration 19300/50000 Total steps: 79,052,800 Steps per second: 11,903 Iteration time: 0.344s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0858 Mean episode length: 1353.9 Actor loss: -9.4671 Critic loss: +0.0035 World model loss: +0.7699 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8340 Episode_Reward/track_angular_velocity: +51.1938 Episode_Reward/upright: +26.0469 Episode_Reward/leg_posture: -2.8174 Episode_Reward/standstill: -6.9196 Episode_Reward/action_rate_l2: -0.5809 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.1187 Episode_Reward/rolling_leg_motion: -0.7085 Episode_Reward/obstacle_progress: +3.0177 Episode_Reward/wheel_body_intrusion: -0.2708 Episode_Reward/pair_body_intrusion: -0.3511 Curriculum/terrain_level_mean: +3.1216 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4832 Episode_Termination/fell_over: 0.0403 Episode_Termination/course_exit: 0.4631 Episode_Termination/illegal_nonwheel_contact: 0.0268 ============================================================================== ============================================================================== Learning iteration 19350/50000 Total steps: 79,257,600 Steps per second: 11,911 Iteration time: 0.344s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1386.2 Actor loss: -9.4775 Critic loss: +0.0037 World model loss: +0.7033 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9855 Episode_Reward/track_angular_velocity: +52.4736 Episode_Reward/upright: +26.6643 Episode_Reward/leg_posture: -2.8349 Episode_Reward/standstill: -6.9268 Episode_Reward/action_rate_l2: -0.6100 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0054 Episode_Reward/unnecessary_wheel_lift: -0.0984 Episode_Reward/rolling_leg_motion: -0.7222 Episode_Reward/obstacle_progress: +3.1081 Episode_Reward/wheel_body_intrusion: -0.2605 Episode_Reward/pair_body_intrusion: -0.3331 Curriculum/terrain_level_mean: +3.1272 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5474 Episode_Termination/fell_over: 0.0146 Episode_Termination/course_exit: 0.4380 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 19400/50000 Total steps: 79,462,400 Steps per second: 11,918 Iteration time: 0.344s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1361.5 Actor loss: -9.5019 Critic loss: +0.0036 World model loss: +0.7241 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0344 Episode_Reward/track_angular_velocity: +51.4718 Episode_Reward/upright: +26.1586 Episode_Reward/leg_posture: -2.7868 Episode_Reward/standstill: -6.4495 Episode_Reward/action_rate_l2: -0.6060 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.1107 Episode_Reward/rolling_leg_motion: -0.7160 Episode_Reward/obstacle_progress: +3.0526 Episode_Reward/wheel_body_intrusion: -0.2502 Episode_Reward/pair_body_intrusion: -0.3105 Curriculum/terrain_level_mean: +3.1345 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0954 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5065 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4740 Episode_Termination/illegal_nonwheel_contact: 0.0195 ============================================================================== ============================================================================== Learning iteration 19450/50000 Total steps: 79,667,200 Steps per second: 11,926 Iteration time: 0.343s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1373.1 Actor loss: -9.4435 Critic loss: +0.0061 World model loss: +0.7375 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4446 Episode_Reward/track_angular_velocity: +51.8186 Episode_Reward/upright: +26.3904 Episode_Reward/leg_posture: -2.7850 Episode_Reward/standstill: -6.3498 Episode_Reward/action_rate_l2: -0.6131 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.1201 Episode_Reward/rolling_leg_motion: -0.7162 Episode_Reward/obstacle_progress: +3.1162 Episode_Reward/wheel_body_intrusion: -0.2558 Episode_Reward/pair_body_intrusion: -0.3313 Curriculum/terrain_level_mean: +3.1392 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0954 Metrics/base_velocity/error_vel_yaw: 0.1375 Episode_Termination/time_out: 0.5153 Episode_Termination/fell_over: 0.0368 Episode_Termination/course_exit: 0.4110 Episode_Termination/illegal_nonwheel_contact: 0.0613 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2364] RECOVERY reset 1 invalid environments; first IDs=[2189] ============================================================================== Learning iteration 19500/50000 Total steps: 79,872,000 Steps per second: 11,933 Iteration time: 0.343s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0863 Mean episode length: 1392.8 Actor loss: -9.4733 Critic loss: +0.0054 World model loss: +0.7814 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0815 Episode_Reward/track_angular_velocity: +52.6420 Episode_Reward/upright: +26.7859 Episode_Reward/leg_posture: -2.8318 Episode_Reward/standstill: -6.4097 Episode_Reward/action_rate_l2: -0.6270 Episode_Reward/joint_torques_l2: -0.1526 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0136 Episode_Reward/unnecessary_wheel_lift: -0.1140 Episode_Reward/rolling_leg_motion: -0.7331 Episode_Reward/obstacle_progress: +3.1287 Episode_Reward/wheel_body_intrusion: -0.2579 Episode_Reward/pair_body_intrusion: -0.3535 Curriculum/terrain_level_mean: +3.1433 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0955 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.5109 Episode_Termination/fell_over: 0.0219 Episode_Termination/course_exit: 0.4526 Episode_Termination/illegal_nonwheel_contact: 0.0219 ============================================================================== ============================================================================== Learning iteration 19550/50000 Total steps: 80,076,800 Steps per second: 11,939 Iteration time: 0.343s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0865 Mean episode length: 1361.7 Actor loss: -9.4550 Critic loss: +0.0040 World model loss: +0.7658 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9816 Episode_Reward/track_angular_velocity: +51.4943 Episode_Reward/upright: +26.1898 Episode_Reward/leg_posture: -2.8125 Episode_Reward/standstill: -6.0449 Episode_Reward/action_rate_l2: -0.5995 Episode_Reward/joint_torques_l2: -0.1511 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0239 Episode_Reward/unnecessary_wheel_lift: -0.1235 Episode_Reward/rolling_leg_motion: -0.6873 Episode_Reward/obstacle_progress: +3.1212 Episode_Reward/wheel_body_intrusion: -0.2631 Episode_Reward/pair_body_intrusion: -0.3303 Curriculum/terrain_level_mean: +3.1479 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4643 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.4714 Episode_Termination/illegal_nonwheel_contact: 0.0500 ============================================================================== ============================================================================== Learning iteration 19600/50000 Total steps: 80,281,600 Steps per second: 11,947 Iteration time: 0.343s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1378.3 Actor loss: -9.4593 Critic loss: +0.0042 World model loss: +0.7710 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4784 Episode_Reward/track_angular_velocity: +52.1094 Episode_Reward/upright: +26.5043 Episode_Reward/leg_posture: -2.8509 Episode_Reward/standstill: -6.4428 Episode_Reward/action_rate_l2: -0.6014 Episode_Reward/joint_torques_l2: -0.1516 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.0941 Episode_Reward/rolling_leg_motion: -0.7067 Episode_Reward/obstacle_progress: +3.1488 Episode_Reward/wheel_body_intrusion: -0.2827 Episode_Reward/pair_body_intrusion: -0.3712 Curriculum/terrain_level_mean: +3.1545 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5208 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.4514 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1370] RECOVERY reset 1 invalid environments; first IDs=[2400] ============================================================================== Learning iteration 19650/50000 Total steps: 80,486,400 Steps per second: 11,954 Iteration time: 0.343s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0864 Mean episode length: 1357.8 Actor loss: -9.4411 Critic loss: +0.0093 World model loss: +0.7410 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8491 Episode_Reward/track_angular_velocity: +51.2912 Episode_Reward/upright: +26.0749 Episode_Reward/leg_posture: -2.8013 Episode_Reward/standstill: -6.0588 Episode_Reward/action_rate_l2: -0.6034 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.1014 Episode_Reward/rolling_leg_motion: -0.6895 Episode_Reward/obstacle_progress: +3.1251 Episode_Reward/wheel_body_intrusion: -0.2779 Episode_Reward/pair_body_intrusion: -0.3561 Curriculum/terrain_level_mean: +3.1614 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0959 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5031 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.4479 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 19700/50000 Total steps: 80,691,200 Steps per second: 11,961 Iteration time: 0.342s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0864 Mean episode length: 1423.4 Actor loss: -9.4739 Critic loss: +0.0038 World model loss: +0.7209 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1065 Episode_Reward/track_angular_velocity: +53.7727 Episode_Reward/upright: +27.3750 Episode_Reward/leg_posture: -2.9087 Episode_Reward/standstill: -6.4197 Episode_Reward/action_rate_l2: -0.6280 Episode_Reward/joint_torques_l2: -0.1570 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0195 Episode_Reward/unnecessary_wheel_lift: -0.1255 Episode_Reward/rolling_leg_motion: -0.7236 Episode_Reward/obstacle_progress: +3.2677 Episode_Reward/wheel_body_intrusion: -0.2701 Episode_Reward/pair_body_intrusion: -0.3415 Curriculum/terrain_level_mean: +3.1658 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.5732 Episode_Termination/fell_over: 0.0122 Episode_Termination/course_exit: 0.3841 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 19750/50000 Total steps: 80,896,000 Steps per second: 11,968 Iteration time: 0.342s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0865 Mean episode length: 1372.5 Actor loss: -9.4729 Critic loss: +0.0048 World model loss: +0.7221 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4033 Episode_Reward/track_angular_velocity: +51.8726 Episode_Reward/upright: +26.4085 Episode_Reward/leg_posture: -2.8307 Episode_Reward/standstill: -6.0127 Episode_Reward/action_rate_l2: -0.6120 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.1106 Episode_Reward/rolling_leg_motion: -0.7119 Episode_Reward/obstacle_progress: +3.0916 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.3406 Curriculum/terrain_level_mean: +3.1707 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4828 Episode_Termination/fell_over: 0.0414 Episode_Termination/course_exit: 0.4552 Episode_Termination/illegal_nonwheel_contact: 0.0552 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2145] RECOVERY reset 1 invalid environments; first IDs=[1370] ============================================================================== Learning iteration 19800/50000 Total steps: 81,100,800 Steps per second: 11,976 Iteration time: 0.342s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1355.1 Actor loss: -9.4584 Critic loss: +0.0039 World model loss: +0.7159 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7358 Episode_Reward/track_angular_velocity: +51.1895 Episode_Reward/upright: +26.0242 Episode_Reward/leg_posture: -2.7884 Episode_Reward/standstill: -6.2654 Episode_Reward/action_rate_l2: -0.6013 Episode_Reward/joint_torques_l2: -0.1498 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.1045 Episode_Reward/rolling_leg_motion: -0.7289 Episode_Reward/obstacle_progress: +2.9627 Episode_Reward/wheel_body_intrusion: -0.2727 Episode_Reward/pair_body_intrusion: -0.3170 Curriculum/terrain_level_mean: +3.1755 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4759 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.4897 Episode_Termination/illegal_nonwheel_contact: 0.0276 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2285] ============================================================================== Learning iteration 19850/50000 Total steps: 81,305,600 Steps per second: 11,983 Iteration time: 0.342s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0861 Mean episode length: 1372.1 Actor loss: -9.4406 Critic loss: +0.0057 World model loss: +0.7461 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5071 Episode_Reward/track_angular_velocity: +51.8670 Episode_Reward/upright: +26.3668 Episode_Reward/leg_posture: -2.8375 Episode_Reward/standstill: -6.4472 Episode_Reward/action_rate_l2: -0.6090 Episode_Reward/joint_torques_l2: -0.1507 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0268 Episode_Reward/unnecessary_wheel_lift: -0.1132 Episode_Reward/rolling_leg_motion: -0.7390 Episode_Reward/obstacle_progress: +2.9806 Episode_Reward/wheel_body_intrusion: -0.2779 Episode_Reward/pair_body_intrusion: -0.3264 Curriculum/terrain_level_mean: +3.1790 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.5152 Episode_Termination/fell_over: 0.0364 Episode_Termination/course_exit: 0.4242 Episode_Termination/illegal_nonwheel_contact: 0.0545 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1744] ============================================================================== Learning iteration 19900/50000 Total steps: 81,510,400 Steps per second: 11,990 Iteration time: 0.342s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0859 Mean episode length: 1350.3 Actor loss: -9.4623 Critic loss: +0.0086 World model loss: +0.7000 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5339 Episode_Reward/track_angular_velocity: +51.0048 Episode_Reward/upright: +25.9358 Episode_Reward/leg_posture: -2.7835 Episode_Reward/standstill: -6.4767 Episode_Reward/action_rate_l2: -0.5991 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0970 Episode_Reward/rolling_leg_motion: -0.7249 Episode_Reward/obstacle_progress: +2.9535 Episode_Reward/wheel_body_intrusion: -0.2644 Episode_Reward/pair_body_intrusion: -0.3595 Curriculum/terrain_level_mean: +3.1880 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4800 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4933 Episode_Termination/illegal_nonwheel_contact: 0.0267 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2264] ============================================================================== Learning iteration 19950/50000 Total steps: 81,715,200 Steps per second: 11,997 Iteration time: 0.341s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1356.6 Actor loss: -9.4730 Critic loss: +0.0037 World model loss: +0.7324 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7676 Episode_Reward/track_angular_velocity: +51.2648 Episode_Reward/upright: +26.0786 Episode_Reward/leg_posture: -2.7799 Episode_Reward/standstill: -6.2708 Episode_Reward/action_rate_l2: -0.6083 Episode_Reward/joint_torques_l2: -0.1482 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0991 Episode_Reward/rolling_leg_motion: -0.7248 Episode_Reward/obstacle_progress: +2.9822 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.3056 Curriculum/terrain_level_mean: +3.1914 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0207 Episode_Termination/course_exit: 0.4414 Episode_Termination/illegal_nonwheel_contact: 0.0483 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1060] ============================================================================== Learning iteration 20000/50000 Total steps: 81,920,000 Steps per second: 12,004 Iteration time: 0.341s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0861 Mean episode length: 1380.1 Actor loss: -9.4767 Critic loss: +0.0052 World model loss: +0.7248 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6078 Episode_Reward/track_angular_velocity: +52.1197 Episode_Reward/upright: +26.4973 Episode_Reward/leg_posture: -2.8599 Episode_Reward/standstill: -6.2791 Episode_Reward/action_rate_l2: -0.6178 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.1001 Episode_Reward/rolling_leg_motion: -0.7358 Episode_Reward/obstacle_progress: +3.0023 Episode_Reward/wheel_body_intrusion: -0.3045 Episode_Reward/pair_body_intrusion: -0.3511 Curriculum/terrain_level_mean: +3.1948 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0959 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5234 Episode_Termination/fell_over: 0.0234 Episode_Termination/course_exit: 0.4297 Episode_Termination/illegal_nonwheel_contact: 0.0391 ============================================================================== Warning: EPA horizon = 24 isn't large enough. RECOVERY reset 1 invalid environments; first IDs=[2206] ============================================================================== Learning iteration 20050/50000 Total steps: 82,124,800 Steps per second: 12,011 Iteration time: 0.341s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1301.6 Actor loss: -9.4323 Critic loss: +0.0061 World model loss: +0.7460 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8034 Episode_Reward/track_angular_velocity: +49.1452 Episode_Reward/upright: +24.9886 Episode_Reward/leg_posture: -2.6881 Episode_Reward/standstill: -5.9642 Episode_Reward/action_rate_l2: -0.5696 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0203 Episode_Reward/unnecessary_wheel_lift: -0.0961 Episode_Reward/rolling_leg_motion: -0.6788 Episode_Reward/obstacle_progress: +2.8813 Episode_Reward/wheel_body_intrusion: -0.2654 Episode_Reward/pair_body_intrusion: -0.2991 Curriculum/terrain_level_mean: +3.1982 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0962 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4638 Episode_Termination/fell_over: 0.0290 Episode_Termination/course_exit: 0.4783 Episode_Termination/illegal_nonwheel_contact: 0.0435 ============================================================================== ============================================================================== Learning iteration 20100/50000 Total steps: 82,329,600 Steps per second: 12,018 Iteration time: 0.341s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1357.4 Actor loss: -9.4766 Critic loss: +0.0040 World model loss: +0.7786 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8544 Episode_Reward/track_angular_velocity: +51.2945 Episode_Reward/upright: +26.0650 Episode_Reward/leg_posture: -2.7971 Episode_Reward/standstill: -6.1611 Episode_Reward/action_rate_l2: -0.5920 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.1049 Episode_Reward/rolling_leg_motion: -0.7149 Episode_Reward/obstacle_progress: +2.9494 Episode_Reward/wheel_body_intrusion: -0.2521 Episode_Reward/pair_body_intrusion: -0.2903 Curriculum/terrain_level_mean: +3.2039 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.4631 Episode_Termination/illegal_nonwheel_contact: 0.0134 ============================================================================== ============================================================================== Learning iteration 20150/50000 Total steps: 82,534,400 Steps per second: 12,025 Iteration time: 0.341s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0866 Mean episode length: 1370.5 Actor loss: -9.4942 Critic loss: +0.0035 World model loss: +0.7186 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3439 Episode_Reward/track_angular_velocity: +51.8315 Episode_Reward/upright: +26.3377 Episode_Reward/leg_posture: -2.7292 Episode_Reward/standstill: -6.0097 Episode_Reward/action_rate_l2: -0.6079 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.1089 Episode_Reward/rolling_leg_motion: -0.7157 Episode_Reward/obstacle_progress: +3.0633 Episode_Reward/wheel_body_intrusion: -0.2500 Episode_Reward/pair_body_intrusion: -0.2952 Curriculum/terrain_level_mean: +3.2090 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4894 Episode_Termination/fell_over: 0.0142 Episode_Termination/course_exit: 0.4752 Episode_Termination/illegal_nonwheel_contact: 0.0355 ============================================================================== ============================================================================== Learning iteration 20200/50000 Total steps: 82,739,200 Steps per second: 12,032 Iteration time: 0.340s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1304.3 Actor loss: -9.4340 Critic loss: +0.0080 World model loss: +0.7725 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8938 Episode_Reward/track_angular_velocity: +49.2847 Episode_Reward/upright: +25.0342 Episode_Reward/leg_posture: -2.6348 Episode_Reward/standstill: -5.9855 Episode_Reward/action_rate_l2: -0.5762 Episode_Reward/joint_torques_l2: -0.1430 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0229 Episode_Reward/unnecessary_wheel_lift: -0.0903 Episode_Reward/rolling_leg_motion: -0.6584 Episode_Reward/obstacle_progress: +2.9477 Episode_Reward/wheel_body_intrusion: -0.2501 Episode_Reward/pair_body_intrusion: -0.3011 Curriculum/terrain_level_mean: +3.2146 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0965 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4130 Episode_Termination/fell_over: 0.0435 Episode_Termination/course_exit: 0.5217 Episode_Termination/illegal_nonwheel_contact: 0.0507 ============================================================================== ============================================================================== Learning iteration 20250/50000 Total steps: 82,944,000 Steps per second: 12,039 Iteration time: 0.340s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0863 Mean episode length: 1413.8 Actor loss: -9.4710 Critic loss: +0.0047 World model loss: +0.7244 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6798 Episode_Reward/track_angular_velocity: +53.4104 Episode_Reward/upright: +27.1662 Episode_Reward/leg_posture: -2.8667 Episode_Reward/standstill: -6.3642 Episode_Reward/action_rate_l2: -0.6411 Episode_Reward/joint_torques_l2: -0.1576 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.1288 Episode_Reward/rolling_leg_motion: -0.7248 Episode_Reward/obstacle_progress: +3.2376 Episode_Reward/wheel_body_intrusion: -0.2755 Episode_Reward/pair_body_intrusion: -0.3142 Curriculum/terrain_level_mean: +3.2202 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0964 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5361 Episode_Termination/fell_over: 0.0181 Episode_Termination/course_exit: 0.4277 Episode_Termination/illegal_nonwheel_contact: 0.0301 ============================================================================== ============================================================================== Learning iteration 20300/50000 Total steps: 83,148,800 Steps per second: 12,046 Iteration time: 0.340s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1357.3 Actor loss: -9.4876 Critic loss: +0.0038 World model loss: +0.7246 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8822 Episode_Reward/track_angular_velocity: +51.3061 Episode_Reward/upright: +26.0410 Episode_Reward/leg_posture: -2.7334 Episode_Reward/standstill: -6.5907 Episode_Reward/action_rate_l2: -0.6062 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0292 Episode_Reward/unnecessary_wheel_lift: -0.1277 Episode_Reward/rolling_leg_motion: -0.6807 Episode_Reward/obstacle_progress: +3.2019 Episode_Reward/wheel_body_intrusion: -0.2688 Episode_Reward/pair_body_intrusion: -0.2927 Curriculum/terrain_level_mean: +3.2241 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5195 Episode_Termination/fell_over: 0.0325 Episode_Termination/course_exit: 0.4091 Episode_Termination/illegal_nonwheel_contact: 0.0519 ============================================================================== ============================================================================== Learning iteration 20350/50000 Total steps: 83,353,600 Steps per second: 12,053 Iteration time: 0.340s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0861 Mean episode length: 1400.4 Actor loss: -9.4819 Critic loss: +0.0041 World model loss: +0.7828 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3339 Episode_Reward/track_angular_velocity: +52.8614 Episode_Reward/upright: +26.8704 Episode_Reward/leg_posture: -2.8515 Episode_Reward/standstill: -6.7084 Episode_Reward/action_rate_l2: -0.6027 Episode_Reward/joint_torques_l2: -0.1539 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0246 Episode_Reward/unnecessary_wheel_lift: -0.1034 Episode_Reward/rolling_leg_motion: -0.6979 Episode_Reward/obstacle_progress: +3.2627 Episode_Reward/wheel_body_intrusion: -0.2829 Episode_Reward/pair_body_intrusion: -0.3263 Curriculum/terrain_level_mean: +3.2266 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 20400/50000 Total steps: 83,558,400 Steps per second: 12,060 Iteration time: 0.340s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0860 Mean episode length: 1338.1 Actor loss: -9.4980 Critic loss: +0.0045 World model loss: +0.7846 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1142 Episode_Reward/track_angular_velocity: +50.6016 Episode_Reward/upright: +25.6829 Episode_Reward/leg_posture: -2.7824 Episode_Reward/standstill: -6.3544 Episode_Reward/action_rate_l2: -0.5764 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0366 Episode_Reward/unnecessary_wheel_lift: -0.1016 Episode_Reward/rolling_leg_motion: -0.6762 Episode_Reward/obstacle_progress: +3.0181 Episode_Reward/wheel_body_intrusion: -0.2908 Episode_Reward/pair_body_intrusion: -0.3278 Curriculum/terrain_level_mean: +3.2295 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4238 Episode_Termination/fell_over: 0.0662 Episode_Termination/course_exit: 0.4834 Episode_Termination/illegal_nonwheel_contact: 0.0596 ============================================================================== ============================================================================== Learning iteration 20450/50000 Total steps: 83,763,200 Steps per second: 12,067 Iteration time: 0.339s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0860 Mean episode length: 1393.3 Actor loss: -9.4483 Critic loss: +0.0074 World model loss: +0.7243 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0649 Episode_Reward/track_angular_velocity: +52.6506 Episode_Reward/upright: +26.7454 Episode_Reward/leg_posture: -2.8985 Episode_Reward/standstill: -6.6500 Episode_Reward/action_rate_l2: -0.5957 Episode_Reward/joint_torques_l2: -0.1534 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.1032 Episode_Reward/rolling_leg_motion: -0.6901 Episode_Reward/obstacle_progress: +3.1161 Episode_Reward/wheel_body_intrusion: -0.2988 Episode_Reward/pair_body_intrusion: -0.3217 Curriculum/terrain_level_mean: +3.2322 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0958 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5076 Episode_Termination/fell_over: 0.0455 Episode_Termination/course_exit: 0.4318 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== ============================================================================== Learning iteration 20500/50000 Total steps: 83,968,000 Steps per second: 12,073 Iteration time: 0.339s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1421.9 Actor loss: -9.4688 Critic loss: +0.0050 World model loss: +0.7166 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1456 Episode_Reward/track_angular_velocity: +53.7459 Episode_Reward/upright: +27.2602 Episode_Reward/leg_posture: -2.9208 Episode_Reward/standstill: -6.9801 Episode_Reward/action_rate_l2: -0.6113 Episode_Reward/joint_torques_l2: -0.1563 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0250 Episode_Reward/unnecessary_wheel_lift: -0.1466 Episode_Reward/rolling_leg_motion: -0.7135 Episode_Reward/obstacle_progress: +3.2055 Episode_Reward/wheel_body_intrusion: -0.2790 Episode_Reward/pair_body_intrusion: -0.2949 Curriculum/terrain_level_mean: +3.2380 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0956 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4968 Episode_Termination/fell_over: 0.0382 Episode_Termination/course_exit: 0.4459 Episode_Termination/illegal_nonwheel_contact: 0.0446 ============================================================================== ============================================================================== Learning iteration 20550/50000 Total steps: 84,172,800 Steps per second: 12,080 Iteration time: 0.339s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0862 Mean episode length: 1391.3 Actor loss: -9.4818 Critic loss: +0.0051 World model loss: +0.7316 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1373 Episode_Reward/track_angular_velocity: +52.6051 Episode_Reward/upright: +26.7143 Episode_Reward/leg_posture: -2.8101 Episode_Reward/standstill: -6.5968 Episode_Reward/action_rate_l2: -0.6115 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0163 Episode_Reward/unnecessary_wheel_lift: -0.1110 Episode_Reward/rolling_leg_motion: -0.7196 Episode_Reward/obstacle_progress: +3.1274 Episode_Reward/wheel_body_intrusion: -0.2469 Episode_Reward/pair_body_intrusion: -0.3248 Curriculum/terrain_level_mean: +3.2419 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0062 Episode_Termination/course_exit: 0.4630 Episode_Termination/illegal_nonwheel_contact: 0.0309 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 20600/50000 Total steps: 84,377,600 Steps per second: 12,087 Iteration time: 0.339s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0866 Mean episode length: 1359.6 Actor loss: -9.4800 Critic loss: +0.0043 World model loss: +0.7249 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0644 Episode_Reward/track_angular_velocity: +51.3988 Episode_Reward/upright: +26.1006 Episode_Reward/leg_posture: -2.7451 Episode_Reward/standstill: -6.1146 Episode_Reward/action_rate_l2: -0.5901 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0239 Episode_Reward/unnecessary_wheel_lift: -0.0950 Episode_Reward/rolling_leg_motion: -0.6846 Episode_Reward/obstacle_progress: +3.1264 Episode_Reward/wheel_body_intrusion: -0.2469 Episode_Reward/pair_body_intrusion: -0.3380 Curriculum/terrain_level_mean: +3.2446 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4264 Episode_Termination/fell_over: 0.0078 Episode_Termination/course_exit: 0.5039 Episode_Termination/illegal_nonwheel_contact: 0.0620 ============================================================================== ============================================================================== Learning iteration 20650/50000 Total steps: 84,582,400 Steps per second: 12,094 Iteration time: 0.339s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1352.0 Actor loss: -9.4516 Critic loss: +0.0102 World model loss: +0.7716 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5999 Episode_Reward/track_angular_velocity: +51.0535 Episode_Reward/upright: +25.9258 Episode_Reward/leg_posture: -2.7262 Episode_Reward/standstill: -6.1834 Episode_Reward/action_rate_l2: -0.5909 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.0976 Episode_Reward/rolling_leg_motion: -0.6699 Episode_Reward/obstacle_progress: +3.1046 Episode_Reward/wheel_body_intrusion: -0.2663 Episode_Reward/pair_body_intrusion: -0.3309 Curriculum/terrain_level_mean: +3.2500 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0960 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4895 Episode_Termination/fell_over: 0.0350 Episode_Termination/course_exit: 0.4685 Episode_Termination/illegal_nonwheel_contact: 0.0210 ============================================================================== ============================================================================== Learning iteration 20700/50000 Total steps: 84,787,200 Steps per second: 12,100 Iteration time: 0.339s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1381.1 Actor loss: -9.4436 Critic loss: +0.0037 World model loss: +0.7407 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8237 Episode_Reward/track_angular_velocity: +52.2276 Episode_Reward/upright: +26.4830 Episode_Reward/leg_posture: -2.7904 Episode_Reward/standstill: -6.6222 Episode_Reward/action_rate_l2: -0.5949 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0936 Episode_Reward/rolling_leg_motion: -0.6768 Episode_Reward/obstacle_progress: +3.1603 Episode_Reward/wheel_body_intrusion: -0.2456 Episode_Reward/pair_body_intrusion: -0.3574 Curriculum/terrain_level_mean: +3.2556 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0286 Episode_Termination/course_exit: 0.4714 Episode_Termination/illegal_nonwheel_contact: 0.0143 ============================================================================== ============================================================================== Learning iteration 20750/50000 Total steps: 84,992,000 Steps per second: 12,107 Iteration time: 0.338s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0861 Mean episode length: 1333.9 Actor loss: -9.4454 Critic loss: +0.0058 World model loss: +0.7108 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9930 Episode_Reward/track_angular_velocity: +50.4486 Episode_Reward/upright: +25.5670 Episode_Reward/leg_posture: -2.7470 Episode_Reward/standstill: -6.3538 Episode_Reward/action_rate_l2: -0.5800 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.1043 Episode_Reward/rolling_leg_motion: -0.6707 Episode_Reward/obstacle_progress: +2.9786 Episode_Reward/wheel_body_intrusion: -0.2728 Episode_Reward/pair_body_intrusion: -0.3004 Curriculum/terrain_level_mean: +3.2620 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0956 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4699 Episode_Termination/fell_over: 0.0241 Episode_Termination/course_exit: 0.4940 Episode_Termination/illegal_nonwheel_contact: 0.0301 ============================================================================== ============================================================================== Learning iteration 20800/50000 Total steps: 85,196,800 Steps per second: 12,114 Iteration time: 0.338s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0859 Mean episode length: 1361.1 Actor loss: -9.4361 Critic loss: +0.0046 World model loss: +0.7592 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0022 Episode_Reward/track_angular_velocity: +51.5007 Episode_Reward/upright: +26.0753 Episode_Reward/leg_posture: -2.7939 Episode_Reward/standstill: -6.7618 Episode_Reward/action_rate_l2: -0.6046 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.1092 Episode_Reward/rolling_leg_motion: -0.6962 Episode_Reward/obstacle_progress: +3.0006 Episode_Reward/wheel_body_intrusion: -0.2645 Episode_Reward/pair_body_intrusion: -0.3126 Curriculum/terrain_level_mean: +3.2668 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.5270 Episode_Termination/fell_over: 0.0203 Episode_Termination/course_exit: 0.4324 Episode_Termination/illegal_nonwheel_contact: 0.0270 ============================================================================== ============================================================================== Learning iteration 20850/50000 Total steps: 85,401,600 Steps per second: 12,120 Iteration time: 0.338s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0858 Mean episode length: 1342.3 Actor loss: -9.4655 Critic loss: +0.0040 World model loss: +0.7012 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2721 Episode_Reward/track_angular_velocity: +50.7883 Episode_Reward/upright: +25.6843 Episode_Reward/leg_posture: -2.7388 Episode_Reward/standstill: -6.7092 Episode_Reward/action_rate_l2: -0.5930 Episode_Reward/joint_torques_l2: -0.1464 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0236 Episode_Reward/unnecessary_wheel_lift: -0.1083 Episode_Reward/rolling_leg_motion: -0.6790 Episode_Reward/obstacle_progress: +2.9687 Episode_Reward/wheel_body_intrusion: -0.2521 Episode_Reward/pair_body_intrusion: -0.3156 Curriculum/terrain_level_mean: +3.2705 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0956 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4610 Episode_Termination/fell_over: 0.0325 Episode_Termination/course_exit: 0.4805 Episode_Termination/illegal_nonwheel_contact: 0.0455 ============================================================================== ============================================================================== Learning iteration 20900/50000 Total steps: 85,606,400 Steps per second: 12,127 Iteration time: 0.338s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0857 Mean episode length: 1372.1 Actor loss: -9.4456 Critic loss: +0.0078 World model loss: +0.7111 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2762 Episode_Reward/track_angular_velocity: +51.9346 Episode_Reward/upright: +26.2252 Episode_Reward/leg_posture: -2.8236 Episode_Reward/standstill: -6.8476 Episode_Reward/action_rate_l2: -0.6008 Episode_Reward/joint_torques_l2: -0.1494 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0295 Episode_Reward/unnecessary_wheel_lift: -0.1162 Episode_Reward/rolling_leg_motion: -0.7002 Episode_Reward/obstacle_progress: +3.0302 Episode_Reward/wheel_body_intrusion: -0.2679 Episode_Reward/pair_body_intrusion: -0.3073 Curriculum/terrain_level_mean: +3.2739 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.5181 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.4157 Episode_Termination/illegal_nonwheel_contact: 0.0482 ============================================================================== ============================================================================== Learning iteration 20950/50000 Total steps: 85,811,200 Steps per second: 12,134 Iteration time: 0.338s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0862 Mean episode length: 1336.3 Actor loss: -9.4475 Critic loss: +0.0056 World model loss: +0.7617 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1311 Episode_Reward/track_angular_velocity: +50.6161 Episode_Reward/upright: +25.5916 Episode_Reward/leg_posture: -2.7270 Episode_Reward/standstill: -6.3389 Episode_Reward/action_rate_l2: -0.5947 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0418 Episode_Reward/unnecessary_wheel_lift: -0.1091 Episode_Reward/rolling_leg_motion: -0.6779 Episode_Reward/obstacle_progress: +2.9992 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.2934 Curriculum/terrain_level_mean: +3.2771 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.5097 Episode_Termination/fell_over: 0.0645 Episode_Termination/course_exit: 0.3871 Episode_Termination/illegal_nonwheel_contact: 0.0645 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[416] ============================================================================== Learning iteration 21000/50000 Total steps: 86,016,000 Steps per second: 12,139 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0861 Mean episode length: 1455.7 Actor loss: -9.4362 Critic loss: +0.0047 World model loss: +0.7485 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.3357 Episode_Reward/track_angular_velocity: +55.1174 Episode_Reward/upright: +27.9201 Episode_Reward/leg_posture: -2.9976 Episode_Reward/standstill: -6.9788 Episode_Reward/action_rate_l2: -0.6431 Episode_Reward/joint_torques_l2: -0.1566 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0114 Episode_Reward/unnecessary_wheel_lift: -0.1153 Episode_Reward/rolling_leg_motion: -0.7473 Episode_Reward/obstacle_progress: +3.2817 Episode_Reward/wheel_body_intrusion: -0.2901 Episode_Reward/pair_body_intrusion: -0.3610 Curriculum/terrain_level_mean: +3.2815 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0951 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.5894 Episode_Termination/fell_over: 0.0265 Episode_Termination/course_exit: 0.3841 Episode_Termination/illegal_nonwheel_contact: 0.0199 ============================================================================== ============================================================================== Learning iteration 21050/50000 Total steps: 86,220,800 Steps per second: 12,146 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1348.2 Actor loss: -9.4627 Critic loss: +0.0039 World model loss: +0.7142 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7490 Episode_Reward/track_angular_velocity: +51.0764 Episode_Reward/upright: +25.8579 Episode_Reward/leg_posture: -2.7351 Episode_Reward/standstill: -6.0904 Episode_Reward/action_rate_l2: -0.5973 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.0937 Episode_Reward/rolling_leg_motion: -0.6656 Episode_Reward/obstacle_progress: +3.0586 Episode_Reward/wheel_body_intrusion: -0.2544 Episode_Reward/pair_body_intrusion: -0.3035 Curriculum/terrain_level_mean: +3.2869 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.4648 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.4930 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== ============================================================================== Learning iteration 21100/50000 Total steps: 86,425,600 Steps per second: 12,152 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0865 Mean episode length: 1401.2 Actor loss: -9.4403 Critic loss: +0.0086 World model loss: +0.7241 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5075 Episode_Reward/track_angular_velocity: +53.0895 Episode_Reward/upright: +26.8996 Episode_Reward/leg_posture: -2.8557 Episode_Reward/standstill: -6.4794 Episode_Reward/action_rate_l2: -0.6060 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0998 Episode_Reward/rolling_leg_motion: -0.6837 Episode_Reward/obstacle_progress: +3.1843 Episode_Reward/wheel_body_intrusion: -0.2836 Episode_Reward/pair_body_intrusion: -0.3250 Curriculum/terrain_level_mean: +3.2913 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.5366 Episode_Termination/fell_over: 0.0244 Episode_Termination/course_exit: 0.4228 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== ============================================================================== Learning iteration 21150/50000 Total steps: 86,630,400 Steps per second: 12,159 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0862 Mean episode length: 1369.4 Actor loss: -9.4495 Critic loss: +0.0038 World model loss: +0.6962 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2243 Episode_Reward/track_angular_velocity: +51.7846 Episode_Reward/upright: +26.2615 Episode_Reward/leg_posture: -2.7933 Episode_Reward/standstill: -6.3209 Episode_Reward/action_rate_l2: -0.6022 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.1162 Episode_Reward/rolling_leg_motion: -0.6778 Episode_Reward/obstacle_progress: +3.0715 Episode_Reward/wheel_body_intrusion: -0.2743 Episode_Reward/pair_body_intrusion: -0.3058 Curriculum/terrain_level_mean: +3.2996 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0953 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0314 Episode_Termination/course_exit: 0.4591 Episode_Termination/illegal_nonwheel_contact: 0.0252 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1141] RECOVERY reset 1 invalid environments; first IDs=[2512] ============================================================================== Learning iteration 21200/50000 Total steps: 86,835,200 Steps per second: 12,165 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0865 Mean episode length: 1383.7 Actor loss: -9.4090 Critic loss: +0.0180 World model loss: +0.7502 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8863 Episode_Reward/track_angular_velocity: +52.3513 Episode_Reward/upright: +26.5351 Episode_Reward/leg_posture: -2.8185 Episode_Reward/standstill: -6.2036 Episode_Reward/action_rate_l2: -0.6103 Episode_Reward/joint_torques_l2: -0.1497 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0297 Episode_Reward/unnecessary_wheel_lift: -0.1242 Episode_Reward/rolling_leg_motion: -0.7127 Episode_Reward/obstacle_progress: +3.2037 Episode_Reward/wheel_body_intrusion: -0.2722 Episode_Reward/pair_body_intrusion: -0.3552 Curriculum/terrain_level_mean: +3.3044 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4974 Episode_Termination/fell_over: 0.0311 Episode_Termination/course_exit: 0.4456 Episode_Termination/illegal_nonwheel_contact: 0.0518 ============================================================================== ============================================================================== Learning iteration 21250/50000 Total steps: 87,040,000 Steps per second: 12,172 Iteration time: 0.337s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0866 Mean episode length: 1348.7 Actor loss: -9.4638 Critic loss: +0.0041 World model loss: +0.7035 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6194 Episode_Reward/track_angular_velocity: +50.9746 Episode_Reward/upright: +25.8788 Episode_Reward/leg_posture: -2.7360 Episode_Reward/standstill: -5.9071 Episode_Reward/action_rate_l2: -0.5813 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.1376 Episode_Reward/rolling_leg_motion: -0.6731 Episode_Reward/obstacle_progress: +3.1430 Episode_Reward/wheel_body_intrusion: -0.2677 Episode_Reward/pair_body_intrusion: -0.3771 Curriculum/terrain_level_mean: +3.3098 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4355 Episode_Termination/fell_over: 0.0161 Episode_Termination/course_exit: 0.5161 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 21300/50000 Total steps: 87,244,800 Steps per second: 12,178 Iteration time: 0.336s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1421.3 Actor loss: -9.4284 Critic loss: +0.0086 World model loss: +0.7100 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1473 Episode_Reward/track_angular_velocity: +53.7458 Episode_Reward/upright: +27.2488 Episode_Reward/leg_posture: -2.8609 Episode_Reward/standstill: -6.2306 Episode_Reward/action_rate_l2: -0.6247 Episode_Reward/joint_torques_l2: -0.1546 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0167 Episode_Reward/unnecessary_wheel_lift: -0.1071 Episode_Reward/rolling_leg_motion: -0.6923 Episode_Reward/obstacle_progress: +3.4728 Episode_Reward/wheel_body_intrusion: -0.2940 Episode_Reward/pair_body_intrusion: -0.3381 Curriculum/terrain_level_mean: +3.3140 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.5271 Episode_Termination/fell_over: 0.0078 Episode_Termination/course_exit: 0.4264 Episode_Termination/illegal_nonwheel_contact: 0.0465 ============================================================================== ============================================================================== Learning iteration 21350/50000 Total steps: 87,449,600 Steps per second: 12,185 Iteration time: 0.336s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0864 Mean episode length: 1292.3 Actor loss: -9.4809 Critic loss: +0.0059 World model loss: +0.7198 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6411 Episode_Reward/track_angular_velocity: +48.8404 Episode_Reward/upright: +24.7333 Episode_Reward/leg_posture: -2.6070 Episode_Reward/standstill: -6.0488 Episode_Reward/action_rate_l2: -0.5468 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0177 Episode_Reward/unnecessary_wheel_lift: -0.0874 Episode_Reward/rolling_leg_motion: -0.6226 Episode_Reward/obstacle_progress: +3.1569 Episode_Reward/wheel_body_intrusion: -0.2565 Episode_Reward/pair_body_intrusion: -0.3262 Curriculum/terrain_level_mean: +3.3174 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4632 Episode_Termination/fell_over: 0.0147 Episode_Termination/course_exit: 0.4853 Episode_Termination/illegal_nonwheel_contact: 0.0441 ============================================================================== ============================================================================== Learning iteration 21400/50000 Total steps: 87,654,400 Steps per second: 12,191 Iteration time: 0.336s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0864 Mean episode length: 1374.8 Actor loss: -9.4545 Critic loss: +0.0040 World model loss: +0.7303 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5278 Episode_Reward/track_angular_velocity: +51.9462 Episode_Reward/upright: +26.3156 Episode_Reward/leg_posture: -2.7196 Episode_Reward/standstill: -6.3735 Episode_Reward/action_rate_l2: -0.5923 Episode_Reward/joint_torques_l2: -0.1508 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0081 Episode_Reward/unnecessary_wheel_lift: -0.0995 Episode_Reward/rolling_leg_motion: -0.6783 Episode_Reward/obstacle_progress: +3.3134 Episode_Reward/wheel_body_intrusion: -0.2786 Episode_Reward/pair_body_intrusion: -0.3492 Curriculum/terrain_level_mean: +3.3245 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5215 Episode_Termination/fell_over: 0.0184 Episode_Termination/course_exit: 0.4601 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 21450/50000 Total steps: 87,859,200 Steps per second: 12,198 Iteration time: 0.336s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0866 Mean episode length: 1401.6 Actor loss: -9.4533 Critic loss: +0.0079 World model loss: +0.7106 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5716 Episode_Reward/track_angular_velocity: +53.0219 Episode_Reward/upright: +26.8727 Episode_Reward/leg_posture: -2.7452 Episode_Reward/standstill: -6.5119 Episode_Reward/action_rate_l2: -0.6053 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.1055 Episode_Reward/rolling_leg_motion: -0.6838 Episode_Reward/obstacle_progress: +3.3469 Episode_Reward/wheel_body_intrusion: -0.2497 Episode_Reward/pair_body_intrusion: -0.3134 Curriculum/terrain_level_mean: +3.3306 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5556 Episode_Termination/fell_over: 0.0309 Episode_Termination/course_exit: 0.4136 Episode_Termination/illegal_nonwheel_contact: 0.0123 ============================================================================== ============================================================================== Learning iteration 21500/50000 Total steps: 88,064,000 Steps per second: 12,204 Iteration time: 0.336s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1354.6 Actor loss: -9.4580 Critic loss: +0.0116 World model loss: +0.6834 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9629 Episode_Reward/track_angular_velocity: +51.2303 Episode_Reward/upright: +25.9728 Episode_Reward/leg_posture: -2.6065 Episode_Reward/standstill: -5.8553 Episode_Reward/action_rate_l2: -0.6011 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0938 Episode_Reward/rolling_leg_motion: -0.6648 Episode_Reward/obstacle_progress: +3.2435 Episode_Reward/wheel_body_intrusion: -0.2480 Episode_Reward/pair_body_intrusion: -0.3248 Curriculum/terrain_level_mean: +3.3359 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4734 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4970 Episode_Termination/illegal_nonwheel_contact: 0.0296 ============================================================================== ============================================================================== Learning iteration 21550/50000 Total steps: 88,268,800 Steps per second: 12,210 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0867 Mean episode length: 1393.6 Actor loss: -9.4694 Critic loss: +0.0047 World model loss: +0.7113 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2154 Episode_Reward/track_angular_velocity: +52.7161 Episode_Reward/upright: +26.7020 Episode_Reward/leg_posture: -2.7564 Episode_Reward/standstill: -6.2658 Episode_Reward/action_rate_l2: -0.6149 Episode_Reward/joint_torques_l2: -0.1519 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.1018 Episode_Reward/rolling_leg_motion: -0.6905 Episode_Reward/obstacle_progress: +3.3444 Episode_Reward/wheel_body_intrusion: -0.2563 Episode_Reward/pair_body_intrusion: -0.3345 Curriculum/terrain_level_mean: +3.3408 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.4690 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 21600/50000 Total steps: 88,473,600 Steps per second: 12,216 Iteration time: 0.335s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1373.0 Actor loss: -9.4770 Critic loss: +0.0037 World model loss: +0.7091 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5074 Episode_Reward/track_angular_velocity: +51.8778 Episode_Reward/upright: +26.2985 Episode_Reward/leg_posture: -2.7294 Episode_Reward/standstill: -5.9552 Episode_Reward/action_rate_l2: -0.6093 Episode_Reward/joint_torques_l2: -0.1502 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.1117 Episode_Reward/rolling_leg_motion: -0.6892 Episode_Reward/obstacle_progress: +3.2423 Episode_Reward/wheel_body_intrusion: -0.2794 Episode_Reward/pair_body_intrusion: -0.3476 Curriculum/terrain_level_mean: +3.3457 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.4699 Episode_Termination/illegal_nonwheel_contact: 0.0181 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_021600.mp4 ============================================================================== Learning iteration 21650/50000 Total steps: 88,678,400 Steps per second: 12,212 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0865 Mean episode length: 1448.6 Actor loss: -9.4543 Critic loss: +0.0100 World model loss: +0.7342 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.1243 Episode_Reward/track_angular_velocity: +54.7020 Episode_Reward/upright: +27.7672 Episode_Reward/leg_posture: -2.9255 Episode_Reward/standstill: -6.4043 Episode_Reward/action_rate_l2: -0.6434 Episode_Reward/joint_torques_l2: -0.1607 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0170 Episode_Reward/unnecessary_wheel_lift: -0.1025 Episode_Reward/rolling_leg_motion: -0.7175 Episode_Reward/obstacle_progress: +3.4180 Episode_Reward/wheel_body_intrusion: -0.3117 Episode_Reward/pair_body_intrusion: -0.4050 Curriculum/terrain_level_mean: +3.3499 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5725 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.3841 Episode_Termination/illegal_nonwheel_contact: 0.0435 ============================================================================== ============================================================================== Learning iteration 21700/50000 Total steps: 88,883,200 Steps per second: 12,219 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1360.9 Actor loss: -9.4621 Critic loss: +0.0107 World model loss: +0.7552 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0044 Episode_Reward/track_angular_velocity: +51.3728 Episode_Reward/upright: +26.0706 Episode_Reward/leg_posture: -2.7297 Episode_Reward/standstill: -5.7691 Episode_Reward/action_rate_l2: -0.5967 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.1156 Episode_Reward/rolling_leg_motion: -0.6597 Episode_Reward/obstacle_progress: +3.2807 Episode_Reward/wheel_body_intrusion: -0.2611 Episode_Reward/pair_body_intrusion: -0.3684 Curriculum/terrain_level_mean: +3.3540 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0951 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.5099 Episode_Termination/fell_over: 0.0199 Episode_Termination/course_exit: 0.4503 Episode_Termination/illegal_nonwheel_contact: 0.0331 ============================================================================== ============================================================================== Learning iteration 21750/50000 Total steps: 89,088,000 Steps per second: 12,225 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1323.7 Actor loss: -9.4930 Critic loss: +0.0037 World model loss: +0.7616 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5381 Episode_Reward/track_angular_velocity: +50.0465 Episode_Reward/upright: +25.3275 Episode_Reward/leg_posture: -2.6392 Episode_Reward/standstill: -6.0614 Episode_Reward/action_rate_l2: -0.5829 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.1138 Episode_Reward/rolling_leg_motion: -0.6416 Episode_Reward/obstacle_progress: +3.1704 Episode_Reward/wheel_body_intrusion: -0.2448 Episode_Reward/pair_body_intrusion: -0.3378 Curriculum/terrain_level_mean: +3.3582 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4348 Episode_Termination/fell_over: 0.0497 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0435 ============================================================================== ============================================================================== Learning iteration 21800/50000 Total steps: 89,292,800 Steps per second: 12,231 Iteration time: 0.335s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0860 Mean episode length: 1339.5 Actor loss: -9.4491 Critic loss: +0.0044 World model loss: +0.7006 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0708 Episode_Reward/track_angular_velocity: +50.5615 Episode_Reward/upright: +25.6237 Episode_Reward/leg_posture: -2.6954 Episode_Reward/standstill: -6.4850 Episode_Reward/action_rate_l2: -0.5829 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0288 Episode_Reward/unnecessary_wheel_lift: -0.1146 Episode_Reward/rolling_leg_motion: -0.6491 Episode_Reward/obstacle_progress: +3.1961 Episode_Reward/wheel_body_intrusion: -0.2626 Episode_Reward/pair_body_intrusion: -0.3245 Curriculum/terrain_level_mean: +3.3616 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0968 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4167 Episode_Termination/fell_over: 0.0606 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0455 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 21850/50000 Total steps: 89,497,600 Steps per second: 12,237 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0863 Mean episode length: 1375.2 Actor loss: -9.4637 Critic loss: +0.0070 World model loss: +0.7398 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4995 Episode_Reward/track_angular_velocity: +51.9871 Episode_Reward/upright: +26.2869 Episode_Reward/leg_posture: -2.7239 Episode_Reward/standstill: -6.5751 Episode_Reward/action_rate_l2: -0.6032 Episode_Reward/joint_torques_l2: -0.1533 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.1040 Episode_Reward/rolling_leg_motion: -0.6704 Episode_Reward/obstacle_progress: +3.3578 Episode_Reward/wheel_body_intrusion: -0.2685 Episode_Reward/pair_body_intrusion: -0.2856 Curriculum/terrain_level_mean: +3.3679 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0951 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4904 Episode_Termination/fell_over: 0.0127 Episode_Termination/course_exit: 0.4777 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[821] RECOVERY reset 1 invalid environments; first IDs=[854] ============================================================================== Learning iteration 21900/50000 Total steps: 89,702,400 Steps per second: 12,244 Iteration time: 0.335s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1394.7 Actor loss: -9.4815 Critic loss: +0.0049 World model loss: +0.7178 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2714 Episode_Reward/track_angular_velocity: +52.7551 Episode_Reward/upright: +26.7257 Episode_Reward/leg_posture: -2.7547 Episode_Reward/standstill: -6.4079 Episode_Reward/action_rate_l2: -0.5881 Episode_Reward/joint_torques_l2: -0.1535 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0857 Episode_Reward/rolling_leg_motion: -0.6741 Episode_Reward/obstacle_progress: +3.3672 Episode_Reward/wheel_body_intrusion: -0.2773 Episode_Reward/pair_body_intrusion: -0.3172 Curriculum/terrain_level_mean: +3.3721 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4793 Episode_Termination/fell_over: 0.0165 Episode_Termination/course_exit: 0.4876 Episode_Termination/illegal_nonwheel_contact: 0.0165 ============================================================================== ============================================================================== Learning iteration 21950/50000 Total steps: 89,907,200 Steps per second: 12,250 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1353.3 Actor loss: -9.4628 Critic loss: +0.0051 World model loss: +0.7589 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7638 Episode_Reward/track_angular_velocity: +51.1093 Episode_Reward/upright: +25.9244 Episode_Reward/leg_posture: -2.6898 Episode_Reward/standstill: -5.7359 Episode_Reward/action_rate_l2: -0.5846 Episode_Reward/joint_torques_l2: -0.1510 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0972 Episode_Reward/rolling_leg_motion: -0.6515 Episode_Reward/obstacle_progress: +3.2829 Episode_Reward/wheel_body_intrusion: -0.2692 Episode_Reward/pair_body_intrusion: -0.3202 Curriculum/terrain_level_mean: +3.3782 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4672 Episode_Termination/fell_over: 0.0219 Episode_Termination/course_exit: 0.5036 Episode_Termination/illegal_nonwheel_contact: 0.0219 ============================================================================== ============================================================================== Learning iteration 22000/50000 Total steps: 90,112,000 Steps per second: 12,256 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1411.0 Actor loss: -9.4906 Critic loss: +0.0033 World model loss: +0.7360 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9309 Episode_Reward/track_angular_velocity: +53.2867 Episode_Reward/upright: +27.0868 Episode_Reward/leg_posture: -2.7504 Episode_Reward/standstill: -6.2043 Episode_Reward/action_rate_l2: -0.6187 Episode_Reward/joint_torques_l2: -0.1556 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.1098 Episode_Reward/rolling_leg_motion: -0.6968 Episode_Reward/obstacle_progress: +3.4457 Episode_Reward/wheel_body_intrusion: -0.2644 Episode_Reward/pair_body_intrusion: -0.3195 Curriculum/terrain_level_mean: +3.3821 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5669 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.4013 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 22050/50000 Total steps: 90,316,800 Steps per second: 12,262 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1438.0 Actor loss: -9.4961 Critic loss: +0.0047 World model loss: +0.7369 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0564 Episode_Reward/track_angular_velocity: +54.3389 Episode_Reward/upright: +27.6005 Episode_Reward/leg_posture: -2.8476 Episode_Reward/standstill: -6.4188 Episode_Reward/action_rate_l2: -0.6149 Episode_Reward/joint_torques_l2: -0.1590 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0253 Episode_Reward/unnecessary_wheel_lift: -0.1040 Episode_Reward/rolling_leg_motion: -0.7140 Episode_Reward/obstacle_progress: +3.4682 Episode_Reward/wheel_body_intrusion: -0.2809 Episode_Reward/pair_body_intrusion: -0.3283 Curriculum/terrain_level_mean: +3.3833 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5779 Episode_Termination/fell_over: 0.0455 Episode_Termination/course_exit: 0.3636 Episode_Termination/illegal_nonwheel_contact: 0.0325 ============================================================================== ============================================================================== Learning iteration 22100/50000 Total steps: 90,521,600 Steps per second: 12,268 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1396.4 Actor loss: -9.4575 Critic loss: +0.0065 World model loss: +0.7623 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6012 Episode_Reward/track_angular_velocity: +52.8563 Episode_Reward/upright: +26.8332 Episode_Reward/leg_posture: -2.8254 Episode_Reward/standstill: -6.0766 Episode_Reward/action_rate_l2: -0.5811 Episode_Reward/joint_torques_l2: -0.1552 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0164 Episode_Reward/unnecessary_wheel_lift: -0.0938 Episode_Reward/rolling_leg_motion: -0.6812 Episode_Reward/obstacle_progress: +3.3263 Episode_Reward/wheel_body_intrusion: -0.2707 Episode_Reward/pair_body_intrusion: -0.3058 Curriculum/terrain_level_mean: +3.3872 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4859 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.4718 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== ============================================================================== Learning iteration 22150/50000 Total steps: 90,726,400 Steps per second: 12,274 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1329.1 Actor loss: -9.4724 Critic loss: +0.0066 World model loss: +0.7268 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0397 Episode_Reward/track_angular_velocity: +50.2717 Episode_Reward/upright: +25.5334 Episode_Reward/leg_posture: -2.6670 Episode_Reward/standstill: -5.6861 Episode_Reward/action_rate_l2: -0.5567 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0182 Episode_Reward/unnecessary_wheel_lift: -0.0852 Episode_Reward/rolling_leg_motion: -0.6405 Episode_Reward/obstacle_progress: +3.1381 Episode_Reward/wheel_body_intrusion: -0.2603 Episode_Reward/pair_body_intrusion: -0.3700 Curriculum/terrain_level_mean: +3.3904 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4586 Episode_Termination/fell_over: 0.0318 Episode_Termination/course_exit: 0.4968 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 22200/50000 Total steps: 90,931,200 Steps per second: 12,280 Iteration time: 0.334s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1342.8 Actor loss: -9.5024 Critic loss: +0.0062 World model loss: +0.7310 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5859 Episode_Reward/track_angular_velocity: +50.8353 Episode_Reward/upright: +25.7897 Episode_Reward/leg_posture: -2.6805 Episode_Reward/standstill: -6.1353 Episode_Reward/action_rate_l2: -0.5651 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.1219 Episode_Reward/rolling_leg_motion: -0.6434 Episode_Reward/obstacle_progress: +3.2488 Episode_Reward/wheel_body_intrusion: -0.2546 Episode_Reward/pair_body_intrusion: -0.3297 Curriculum/terrain_level_mean: +3.3950 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4315 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.5137 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 22250/50000 Total steps: 91,136,000 Steps per second: 12,286 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0872 Mean episode length: 1414.4 Actor loss: -9.5025 Critic loss: +0.0032 World model loss: +0.6861 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2260 Episode_Reward/track_angular_velocity: +53.5814 Episode_Reward/upright: +27.1648 Episode_Reward/leg_posture: -2.8220 Episode_Reward/standstill: -5.9639 Episode_Reward/action_rate_l2: -0.6043 Episode_Reward/joint_torques_l2: -0.1551 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0193 Episode_Reward/unnecessary_wheel_lift: -0.1144 Episode_Reward/rolling_leg_motion: -0.7040 Episode_Reward/obstacle_progress: +3.3954 Episode_Reward/wheel_body_intrusion: -0.2827 Episode_Reward/pair_body_intrusion: -0.3634 Curriculum/terrain_level_mean: +3.3992 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.5429 Episode_Termination/fell_over: 0.0286 Episode_Termination/course_exit: 0.4071 Episode_Termination/illegal_nonwheel_contact: 0.0214 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1650] ============================================================================== Learning iteration 22300/50000 Total steps: 91,340,800 Steps per second: 12,292 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1398.6 Actor loss: -9.4995 Critic loss: +0.0037 World model loss: +0.7358 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5154 Episode_Reward/track_angular_velocity: +52.9527 Episode_Reward/upright: +26.8525 Episode_Reward/leg_posture: -2.7510 Episode_Reward/standstill: -5.7701 Episode_Reward/action_rate_l2: -0.6079 Episode_Reward/joint_torques_l2: -0.1544 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.1111 Episode_Reward/rolling_leg_motion: -0.6905 Episode_Reward/obstacle_progress: +3.3810 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.3594 Curriculum/terrain_level_mean: +3.4048 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4939 Episode_Termination/fell_over: 0.0061 Episode_Termination/course_exit: 0.4573 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== ============================================================================== Learning iteration 22350/50000 Total steps: 91,545,600 Steps per second: 12,298 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0871 Mean episode length: 1256.0 Actor loss: -9.4982 Critic loss: +0.0037 World model loss: +0.6739 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.4905 Episode_Reward/track_angular_velocity: +47.4742 Episode_Reward/upright: +24.1014 Episode_Reward/leg_posture: -2.4830 Episode_Reward/standstill: -5.3461 Episode_Reward/action_rate_l2: -0.5434 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0172 Episode_Reward/unnecessary_wheel_lift: -0.0817 Episode_Reward/rolling_leg_motion: -0.6107 Episode_Reward/obstacle_progress: +3.1053 Episode_Reward/wheel_body_intrusion: -0.2399 Episode_Reward/pair_body_intrusion: -0.2939 Curriculum/terrain_level_mean: +3.4097 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.3958 Episode_Termination/fell_over: 0.0347 Episode_Termination/course_exit: 0.5556 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1610] ============================================================================== Learning iteration 22400/50000 Total steps: 91,750,400 Steps per second: 12,303 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1321.3 Actor loss: -9.4907 Critic loss: +0.0075 World model loss: +0.7200 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6520 Episode_Reward/track_angular_velocity: +49.9533 Episode_Reward/upright: +25.3565 Episode_Reward/leg_posture: -2.6552 Episode_Reward/standstill: -5.4491 Episode_Reward/action_rate_l2: -0.5641 Episode_Reward/joint_torques_l2: -0.1477 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0129 Episode_Reward/unnecessary_wheel_lift: -0.0914 Episode_Reward/rolling_leg_motion: -0.6328 Episode_Reward/obstacle_progress: +3.2455 Episode_Reward/wheel_body_intrusion: -0.2599 Episode_Reward/pair_body_intrusion: -0.3104 Curriculum/terrain_level_mean: +3.4172 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0185 Episode_Termination/course_exit: 0.5247 Episode_Termination/illegal_nonwheel_contact: 0.0185 ============================================================================== ============================================================================== Learning iteration 22450/50000 Total steps: 91,955,200 Steps per second: 12,309 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0864 Mean episode length: 1366.5 Actor loss: -9.5070 Critic loss: +0.0035 World model loss: +0.7247 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1598 Episode_Reward/track_angular_velocity: +51.5642 Episode_Reward/upright: +26.1916 Episode_Reward/leg_posture: -2.7559 Episode_Reward/standstill: -6.2711 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1546 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.1079 Episode_Reward/rolling_leg_motion: -0.6683 Episode_Reward/obstacle_progress: +3.3533 Episode_Reward/wheel_body_intrusion: -0.2707 Episode_Reward/pair_body_intrusion: -0.3264 Curriculum/terrain_level_mean: +3.4226 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0956 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.4726 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== ============================================================================== Learning iteration 22500/50000 Total steps: 92,160,000 Steps per second: 12,315 Iteration time: 0.333s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0866 Mean episode length: 1394.5 Actor loss: -9.4938 Critic loss: +0.0035 World model loss: +0.7552 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2035 Episode_Reward/track_angular_velocity: +52.6623 Episode_Reward/upright: +26.7171 Episode_Reward/leg_posture: -2.7655 Episode_Reward/standstill: -6.4213 Episode_Reward/action_rate_l2: -0.6255 Episode_Reward/joint_torques_l2: -0.1566 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0191 Episode_Reward/unnecessary_wheel_lift: -0.1078 Episode_Reward/rolling_leg_motion: -0.6735 Episode_Reward/obstacle_progress: +3.5187 Episode_Reward/wheel_body_intrusion: -0.2442 Episode_Reward/pair_body_intrusion: -0.2955 Curriculum/terrain_level_mean: +3.4275 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0951 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.5062 Episode_Termination/fell_over: 0.0370 Episode_Termination/course_exit: 0.4506 Episode_Termination/illegal_nonwheel_contact: 0.0247 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1580] ============================================================================== Learning iteration 22550/50000 Total steps: 92,364,800 Steps per second: 12,321 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0870 Mean episode length: 1390.4 Actor loss: -9.5290 Critic loss: +0.0055 World model loss: +0.7310 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2091 Episode_Reward/track_angular_velocity: +52.5432 Episode_Reward/upright: +26.6597 Episode_Reward/leg_posture: -2.7547 Episode_Reward/standstill: -6.0420 Episode_Reward/action_rate_l2: -0.5999 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0163 Episode_Reward/unnecessary_wheel_lift: -0.0897 Episode_Reward/rolling_leg_motion: -0.6715 Episode_Reward/obstacle_progress: +3.4668 Episode_Reward/wheel_body_intrusion: -0.2460 Episode_Reward/pair_body_intrusion: -0.2827 Curriculum/terrain_level_mean: +3.4319 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4658 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.4932 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2336] RECOVERY reset 1 invalid environments; first IDs=[2548] ============================================================================== Learning iteration 22600/50000 Total steps: 92,569,600 Steps per second: 12,327 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0867 Mean episode length: 1362.2 Actor loss: -9.5058 Critic loss: +0.0073 World model loss: +0.6907 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0598 Episode_Reward/track_angular_velocity: +51.4760 Episode_Reward/upright: +26.0982 Episode_Reward/leg_posture: -2.7239 Episode_Reward/standstill: -6.1245 Episode_Reward/action_rate_l2: -0.5999 Episode_Reward/joint_torques_l2: -0.1530 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0945 Episode_Reward/rolling_leg_motion: -0.6623 Episode_Reward/obstacle_progress: +3.3946 Episode_Reward/wheel_body_intrusion: -0.2718 Episode_Reward/pair_body_intrusion: -0.3196 Curriculum/terrain_level_mean: +3.4390 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4615 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.5192 Episode_Termination/illegal_nonwheel_contact: 0.0128 ============================================================================== ============================================================================== Learning iteration 22650/50000 Total steps: 92,774,400 Steps per second: 12,333 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0867 Mean episode length: 1400.7 Actor loss: -9.4654 Critic loss: +0.0073 World model loss: +0.7115 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4673 Episode_Reward/track_angular_velocity: +52.9666 Episode_Reward/upright: +26.8751 Episode_Reward/leg_posture: -2.8322 Episode_Reward/standstill: -6.3004 Episode_Reward/action_rate_l2: -0.6158 Episode_Reward/joint_torques_l2: -0.1569 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.1215 Episode_Reward/rolling_leg_motion: -0.6829 Episode_Reward/obstacle_progress: +3.4889 Episode_Reward/wheel_body_intrusion: -0.2816 Episode_Reward/pair_body_intrusion: -0.3174 Curriculum/terrain_level_mean: +3.4429 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4867 Episode_Termination/fell_over: 0.0467 Episode_Termination/course_exit: 0.4600 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== ============================================================================== Learning iteration 22700/50000 Total steps: 92,979,200 Steps per second: 12,339 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1361.8 Actor loss: -9.4883 Critic loss: +0.0046 World model loss: +0.7475 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0857 Episode_Reward/track_angular_velocity: +51.4556 Episode_Reward/upright: +26.1187 Episode_Reward/leg_posture: -2.6880 Episode_Reward/standstill: -5.8545 Episode_Reward/action_rate_l2: -0.5907 Episode_Reward/joint_torques_l2: -0.1510 Episode_Reward/dof_pos_limits: -0.0022 Episode_Reward/terminated: -0.0293 Episode_Reward/unnecessary_wheel_lift: -0.1023 Episode_Reward/rolling_leg_motion: -0.6534 Episode_Reward/obstacle_progress: +3.4553 Episode_Reward/wheel_body_intrusion: -0.2612 Episode_Reward/pair_body_intrusion: -0.3206 Curriculum/terrain_level_mean: +3.4468 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4452 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.4795 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2337] ============================================================================== Learning iteration 22750/50000 Total steps: 93,184,000 Steps per second: 12,345 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1399.0 Actor loss: -9.4851 Critic loss: +0.0052 World model loss: +0.7154 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3028 Episode_Reward/track_angular_velocity: +52.8485 Episode_Reward/upright: +26.8402 Episode_Reward/leg_posture: -2.8031 Episode_Reward/standstill: -6.1015 Episode_Reward/action_rate_l2: -0.6098 Episode_Reward/joint_torques_l2: -0.1544 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0383 Episode_Reward/unnecessary_wheel_lift: -0.1325 Episode_Reward/rolling_leg_motion: -0.6921 Episode_Reward/obstacle_progress: +3.5299 Episode_Reward/wheel_body_intrusion: -0.2700 Episode_Reward/pair_body_intrusion: -0.3334 Curriculum/terrain_level_mean: +3.4485 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0957 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4882 Episode_Termination/fell_over: 0.0294 Episode_Termination/course_exit: 0.4294 Episode_Termination/illegal_nonwheel_contact: 0.0706 ============================================================================== ============================================================================== Learning iteration 22800/50000 Total steps: 93,388,800 Steps per second: 12,351 Iteration time: 0.332s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1341.3 Actor loss: -9.4952 Critic loss: +0.0074 World model loss: +0.7120 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4002 Episode_Reward/track_angular_velocity: +50.7493 Episode_Reward/upright: +25.7502 Episode_Reward/leg_posture: -2.6650 Episode_Reward/standstill: -5.6158 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0236 Episode_Reward/unnecessary_wheel_lift: -0.0952 Episode_Reward/rolling_leg_motion: -0.6554 Episode_Reward/obstacle_progress: +3.2686 Episode_Reward/wheel_body_intrusion: -0.2358 Episode_Reward/pair_body_intrusion: -0.3197 Curriculum/terrain_level_mean: +3.4519 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4476 Episode_Termination/fell_over: 0.0280 Episode_Termination/course_exit: 0.4895 Episode_Termination/illegal_nonwheel_contact: 0.0559 ============================================================================== ============================================================================== Learning iteration 22850/50000 Total steps: 93,593,600 Steps per second: 12,357 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1318.5 Actor loss: -9.5098 Critic loss: +0.0031 World model loss: +0.7052 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6348 Episode_Reward/track_angular_velocity: +49.9145 Episode_Reward/upright: +25.3040 Episode_Reward/leg_posture: -2.5684 Episode_Reward/standstill: -5.3443 Episode_Reward/action_rate_l2: -0.5721 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0896 Episode_Reward/rolling_leg_motion: -0.6267 Episode_Reward/obstacle_progress: +3.3117 Episode_Reward/wheel_body_intrusion: -0.2322 Episode_Reward/pair_body_intrusion: -0.2814 Curriculum/terrain_level_mean: +3.4561 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4565 Episode_Termination/fell_over: 0.0290 Episode_Termination/course_exit: 0.4855 Episode_Termination/illegal_nonwheel_contact: 0.0507 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1872] ============================================================================== Learning iteration 22900/50000 Total steps: 93,798,400 Steps per second: 12,362 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0874 Mean episode length: 1367.3 Actor loss: -9.4847 Critic loss: +0.0063 World model loss: +0.7616 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3229 Episode_Reward/track_angular_velocity: +51.7354 Episode_Reward/upright: +26.2478 Episode_Reward/leg_posture: -2.6555 Episode_Reward/standstill: -5.4069 Episode_Reward/action_rate_l2: -0.5960 Episode_Reward/joint_torques_l2: -0.1506 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.0970 Episode_Reward/rolling_leg_motion: -0.6644 Episode_Reward/obstacle_progress: +3.3839 Episode_Reward/wheel_body_intrusion: -0.2530 Episode_Reward/pair_body_intrusion: -0.2977 Curriculum/terrain_level_mean: +3.4612 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4583 Episode_Termination/fell_over: 0.0238 Episode_Termination/course_exit: 0.4940 Episode_Termination/illegal_nonwheel_contact: 0.0298 ============================================================================== ============================================================================== Learning iteration 22950/50000 Total steps: 94,003,200 Steps per second: 12,368 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1378.5 Actor loss: -9.4962 Critic loss: +0.0033 World model loss: +0.7193 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5810 Episode_Reward/track_angular_velocity: +52.1379 Episode_Reward/upright: +26.4658 Episode_Reward/leg_posture: -2.6945 Episode_Reward/standstill: -5.5464 Episode_Reward/action_rate_l2: -0.6131 Episode_Reward/joint_torques_l2: -0.1519 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.1004 Episode_Reward/rolling_leg_motion: -0.6677 Episode_Reward/obstacle_progress: +3.3358 Episode_Reward/wheel_body_intrusion: -0.2371 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +3.4678 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0954 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4881 Episode_Termination/fell_over: 0.0119 Episode_Termination/course_exit: 0.4762 Episode_Termination/illegal_nonwheel_contact: 0.0298 ============================================================================== ============================================================================== Learning iteration 23000/50000 Total steps: 94,208,000 Steps per second: 12,373 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0871 Mean episode length: 1222.1 Actor loss: -9.5077 Critic loss: +0.0029 World model loss: +0.7191 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +43.1853 Episode_Reward/track_angular_velocity: +46.1783 Episode_Reward/upright: +23.4364 Episode_Reward/leg_posture: -2.3792 Episode_Reward/standstill: -5.1744 Episode_Reward/action_rate_l2: -0.5378 Episode_Reward/joint_torques_l2: -0.1350 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0263 Episode_Reward/unnecessary_wheel_lift: -0.0956 Episode_Reward/rolling_leg_motion: -0.5767 Episode_Reward/obstacle_progress: +3.0614 Episode_Reward/wheel_body_intrusion: -0.2268 Episode_Reward/pair_body_intrusion: -0.2732 Curriculum/terrain_level_mean: +3.4697 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.3493 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.5753 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1447] ============================================================================== Learning iteration 23050/50000 Total steps: 94,412,800 Steps per second: 12,379 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1339.6 Actor loss: -9.4752 Critic loss: +0.0038 World model loss: +0.7191 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1180 Episode_Reward/track_angular_velocity: +50.5632 Episode_Reward/upright: +25.6707 Episode_Reward/leg_posture: -2.5955 Episode_Reward/standstill: -5.8008 Episode_Reward/action_rate_l2: -0.6022 Episode_Reward/joint_torques_l2: -0.1479 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0858 Episode_Reward/rolling_leg_motion: -0.6369 Episode_Reward/obstacle_progress: +3.3011 Episode_Reward/wheel_body_intrusion: -0.2503 Episode_Reward/pair_body_intrusion: -0.3278 Curriculum/terrain_level_mean: +3.4736 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0967 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4207 Episode_Termination/fell_over: 0.0276 Episode_Termination/course_exit: 0.5241 Episode_Termination/illegal_nonwheel_contact: 0.0483 ============================================================================== ============================================================================== Learning iteration 23100/50000 Total steps: 94,617,600 Steps per second: 12,385 Iteration time: 0.331s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1293.4 Actor loss: -9.4962 Critic loss: +0.0040 World model loss: +0.7094 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5381 Episode_Reward/track_angular_velocity: +48.7950 Episode_Reward/upright: +24.7898 Episode_Reward/leg_posture: -2.5099 Episode_Reward/standstill: -5.3483 Episode_Reward/action_rate_l2: -0.5795 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0152 Episode_Reward/unnecessary_wheel_lift: -0.0840 Episode_Reward/rolling_leg_motion: -0.6086 Episode_Reward/obstacle_progress: +3.2497 Episode_Reward/wheel_body_intrusion: -0.2351 Episode_Reward/pair_body_intrusion: -0.3226 Curriculum/terrain_level_mean: +3.4768 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0963 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.4275 Episode_Termination/fell_over: 0.0145 Episode_Termination/course_exit: 0.5290 Episode_Termination/illegal_nonwheel_contact: 0.0362 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2212] ============================================================================== Learning iteration 23150/50000 Total steps: 94,822,400 Steps per second: 12,390 Iteration time: 0.331s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1382.3 Actor loss: -9.4839 Critic loss: +0.0041 World model loss: +0.7002 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5850 Episode_Reward/track_angular_velocity: +52.1016 Episode_Reward/upright: +26.4997 Episode_Reward/leg_posture: -2.6344 Episode_Reward/standstill: -5.8932 Episode_Reward/action_rate_l2: -0.6220 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0297 Episode_Reward/unnecessary_wheel_lift: -0.0996 Episode_Reward/rolling_leg_motion: -0.6469 Episode_Reward/obstacle_progress: +3.5274 Episode_Reward/wheel_body_intrusion: -0.2448 Episode_Reward/pair_body_intrusion: -0.3141 Curriculum/terrain_level_mean: +3.4795 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0971 Metrics/base_velocity/error_vel_yaw: 0.1384 Episode_Termination/time_out: 0.4901 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.4371 Episode_Termination/illegal_nonwheel_contact: 0.0662 ============================================================================== ============================================================================== Learning iteration 23200/50000 Total steps: 95,027,200 Steps per second: 12,396 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1422.2 Actor loss: -9.4758 Critic loss: +0.0052 World model loss: +0.7310 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1238 Episode_Reward/track_angular_velocity: +53.6665 Episode_Reward/upright: +27.2865 Episode_Reward/leg_posture: -2.6636 Episode_Reward/standstill: -5.8437 Episode_Reward/action_rate_l2: -0.6365 Episode_Reward/joint_torques_l2: -0.1553 Episode_Reward/dof_pos_limits: -0.0020 Episode_Reward/terminated: -0.0194 Episode_Reward/unnecessary_wheel_lift: -0.0965 Episode_Reward/rolling_leg_motion: -0.6621 Episode_Reward/obstacle_progress: +3.6169 Episode_Reward/wheel_body_intrusion: -0.2523 Episode_Reward/pair_body_intrusion: -0.3110 Curriculum/terrain_level_mean: +3.4829 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0961 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.5548 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== ============================================================================== Learning iteration 23250/50000 Total steps: 95,232,000 Steps per second: 12,402 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1400.6 Actor loss: -9.4587 Critic loss: +0.0108 World model loss: +0.7242 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4424 Episode_Reward/track_angular_velocity: +52.8350 Episode_Reward/upright: +26.9189 Episode_Reward/leg_posture: -2.6602 Episode_Reward/standstill: -5.8489 Episode_Reward/action_rate_l2: -0.6233 Episode_Reward/joint_torques_l2: -0.1543 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.1094 Episode_Reward/rolling_leg_motion: -0.6699 Episode_Reward/obstacle_progress: +3.6085 Episode_Reward/wheel_body_intrusion: -0.2653 Episode_Reward/pair_body_intrusion: -0.3057 Curriculum/terrain_level_mean: +3.4846 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1376 Episode_Termination/time_out: 0.5127 Episode_Termination/fell_over: 0.0443 Episode_Termination/course_exit: 0.4367 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[891] ============================================================================== Learning iteration 23300/50000 Total steps: 95,436,800 Steps per second: 12,407 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1262.9 Actor loss: -9.5064 Critic loss: +0.0032 World model loss: +0.7390 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.7384 Episode_Reward/track_angular_velocity: +47.7024 Episode_Reward/upright: +24.2935 Episode_Reward/leg_posture: -2.4178 Episode_Reward/standstill: -5.0653 Episode_Reward/action_rate_l2: -0.5595 Episode_Reward/joint_torques_l2: -0.1386 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0123 Episode_Reward/unnecessary_wheel_lift: -0.0916 Episode_Reward/rolling_leg_motion: -0.6010 Episode_Reward/obstacle_progress: +3.2523 Episode_Reward/wheel_body_intrusion: -0.2522 Episode_Reward/pair_body_intrusion: -0.2720 Curriculum/terrain_level_mean: +3.4871 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4361 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.5263 Episode_Termination/illegal_nonwheel_contact: 0.0226 ============================================================================== ============================================================================== Learning iteration 23350/50000 Total steps: 95,641,600 Steps per second: 12,413 Iteration time: 0.330s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1324.0 Actor loss: -9.4700 Critic loss: +0.0040 World model loss: +0.7126 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9420 Episode_Reward/track_angular_velocity: +50.0370 Episode_Reward/upright: +25.4470 Episode_Reward/leg_posture: -2.5156 Episode_Reward/standstill: -5.4133 Episode_Reward/action_rate_l2: -0.5767 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.1115 Episode_Reward/rolling_leg_motion: -0.6379 Episode_Reward/obstacle_progress: +3.3612 Episode_Reward/wheel_body_intrusion: -0.2493 Episode_Reward/pair_body_intrusion: -0.3044 Curriculum/terrain_level_mean: +3.4880 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4671 Episode_Termination/fell_over: 0.0329 Episode_Termination/course_exit: 0.4671 Episode_Termination/illegal_nonwheel_contact: 0.0658 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2018] ============================================================================== Learning iteration 23400/50000 Total steps: 95,846,400 Steps per second: 12,418 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1360.4 Actor loss: -9.4856 Critic loss: +0.0088 World model loss: +0.6721 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2234 Episode_Reward/track_angular_velocity: +51.3643 Episode_Reward/upright: +26.1016 Episode_Reward/leg_posture: -2.5950 Episode_Reward/standstill: -6.1023 Episode_Reward/action_rate_l2: -0.5931 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.0974 Episode_Reward/rolling_leg_motion: -0.6377 Episode_Reward/obstacle_progress: +3.5027 Episode_Reward/wheel_body_intrusion: -0.2610 Episode_Reward/pair_body_intrusion: -0.2821 Curriculum/terrain_level_mean: +3.4910 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4702 Episode_Termination/fell_over: 0.0199 Episode_Termination/course_exit: 0.4901 Episode_Termination/illegal_nonwheel_contact: 0.0199 ============================================================================== ============================================================================== Learning iteration 23450/50000 Total steps: 96,051,200 Steps per second: 12,424 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0871 Mean episode length: 1338.0 Actor loss: -9.5122 Critic loss: +0.0030 World model loss: +0.7431 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3293 Episode_Reward/track_angular_velocity: +50.4853 Episode_Reward/upright: +25.6816 Episode_Reward/leg_posture: -2.5728 Episode_Reward/standstill: -5.9568 Episode_Reward/action_rate_l2: -0.5688 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0261 Episode_Reward/unnecessary_wheel_lift: -0.1023 Episode_Reward/rolling_leg_motion: -0.6204 Episode_Reward/obstacle_progress: +3.5169 Episode_Reward/wheel_body_intrusion: -0.2452 Episode_Reward/pair_body_intrusion: -0.2656 Curriculum/terrain_level_mean: +3.4949 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4459 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4904 Episode_Termination/illegal_nonwheel_contact: 0.0446 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1462] ============================================================================== Learning iteration 23500/50000 Total steps: 96,256,000 Steps per second: 12,430 Iteration time: 0.330s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1354.1 Actor loss: -9.4899 Critic loss: +0.0039 World model loss: +0.7644 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9427 Episode_Reward/track_angular_velocity: +51.1564 Episode_Reward/upright: +26.0205 Episode_Reward/leg_posture: -2.5671 Episode_Reward/standstill: -5.5870 Episode_Reward/action_rate_l2: -0.5813 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0344 Episode_Reward/unnecessary_wheel_lift: -0.1028 Episode_Reward/rolling_leg_motion: -0.6313 Episode_Reward/obstacle_progress: +3.4964 Episode_Reward/wheel_body_intrusion: -0.2436 Episode_Reward/pair_body_intrusion: -0.2411 Curriculum/terrain_level_mean: +3.4961 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5163 Episode_Termination/fell_over: 0.0392 Episode_Termination/course_exit: 0.4052 Episode_Termination/illegal_nonwheel_contact: 0.0719 ============================================================================== ============================================================================== Learning iteration 23550/50000 Total steps: 96,460,800 Steps per second: 12,435 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0879 Mean episode length: 1261.4 Actor loss: -9.5098 Critic loss: +0.0038 World model loss: +0.7224 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.7699 Episode_Reward/track_angular_velocity: +47.6992 Episode_Reward/upright: +24.2376 Episode_Reward/leg_posture: -2.3702 Episode_Reward/standstill: -5.0143 Episode_Reward/action_rate_l2: -0.5377 Episode_Reward/joint_torques_l2: -0.1382 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0222 Episode_Reward/unnecessary_wheel_lift: -0.0960 Episode_Reward/rolling_leg_motion: -0.5893 Episode_Reward/obstacle_progress: +3.3617 Episode_Reward/wheel_body_intrusion: -0.2091 Episode_Reward/pair_body_intrusion: -0.2482 Curriculum/terrain_level_mean: +3.4983 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4091 Episode_Termination/fell_over: 0.0390 Episode_Termination/course_exit: 0.5325 Episode_Termination/illegal_nonwheel_contact: 0.0325 ============================================================================== ============================================================================== Learning iteration 23600/50000 Total steps: 96,665,600 Steps per second: 12,440 Iteration time: 0.329s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1382.2 Actor loss: -9.5264 Critic loss: +0.0063 World model loss: +0.7382 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0778 Episode_Reward/track_angular_velocity: +52.2748 Episode_Reward/upright: +26.5784 Episode_Reward/leg_posture: -2.6220 Episode_Reward/standstill: -5.5844 Episode_Reward/action_rate_l2: -0.6012 Episode_Reward/joint_torques_l2: -0.1518 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0901 Episode_Reward/rolling_leg_motion: -0.6340 Episode_Reward/obstacle_progress: +3.6860 Episode_Reward/wheel_body_intrusion: -0.2581 Episode_Reward/pair_body_intrusion: -0.3021 Curriculum/terrain_level_mean: +3.5024 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5135 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.4459 Episode_Termination/illegal_nonwheel_contact: 0.0135 ============================================================================== ============================================================================== Learning iteration 23650/50000 Total steps: 96,870,400 Steps per second: 12,446 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1300.6 Actor loss: -9.4938 Critic loss: +0.0054 World model loss: +0.6697 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1816 Episode_Reward/track_angular_velocity: +49.2088 Episode_Reward/upright: +24.9945 Episode_Reward/leg_posture: -2.4531 Episode_Reward/standstill: -5.5545 Episode_Reward/action_rate_l2: -0.5590 Episode_Reward/joint_torques_l2: -0.1422 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0178 Episode_Reward/unnecessary_wheel_lift: -0.0881 Episode_Reward/rolling_leg_motion: -0.5800 Episode_Reward/obstacle_progress: +3.4753 Episode_Reward/wheel_body_intrusion: -0.2392 Episode_Reward/pair_body_intrusion: -0.2832 Curriculum/terrain_level_mean: +3.5061 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4029 Episode_Termination/fell_over: 0.0144 Episode_Termination/course_exit: 0.5468 Episode_Termination/illegal_nonwheel_contact: 0.0432 ============================================================================== ============================================================================== Learning iteration 23700/50000 Total steps: 97,075,200 Steps per second: 12,451 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1306.7 Actor loss: -9.5002 Critic loss: +0.0044 World model loss: +0.7113 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4068 Episode_Reward/track_angular_velocity: +49.4388 Episode_Reward/upright: +25.1139 Episode_Reward/leg_posture: -2.4332 Episode_Reward/standstill: -6.0180 Episode_Reward/action_rate_l2: -0.5853 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0306 Episode_Reward/unnecessary_wheel_lift: -0.1095 Episode_Reward/rolling_leg_motion: -0.6034 Episode_Reward/obstacle_progress: +3.4331 Episode_Reward/wheel_body_intrusion: -0.2429 Episode_Reward/pair_body_intrusion: -0.2646 Curriculum/terrain_level_mean: +3.5093 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4337 Episode_Termination/fell_over: 0.0482 Episode_Termination/course_exit: 0.4940 Episode_Termination/illegal_nonwheel_contact: 0.0361 ============================================================================== ============================================================================== Learning iteration 23750/50000 Total steps: 97,280,000 Steps per second: 12,457 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1367.0 Actor loss: -9.5295 Critic loss: +0.0087 World model loss: +0.7454 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5594 Episode_Reward/track_angular_velocity: +51.6949 Episode_Reward/upright: +26.2824 Episode_Reward/leg_posture: -2.5623 Episode_Reward/standstill: -6.2887 Episode_Reward/action_rate_l2: -0.5978 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.1017 Episode_Reward/rolling_leg_motion: -0.6310 Episode_Reward/obstacle_progress: +3.4959 Episode_Reward/wheel_body_intrusion: -0.2515 Episode_Reward/pair_body_intrusion: -0.3277 Curriculum/terrain_level_mean: +3.5132 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4606 Episode_Termination/fell_over: 0.0121 Episode_Termination/course_exit: 0.5030 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== ============================================================================== Learning iteration 23800/50000 Total steps: 97,484,800 Steps per second: 12,462 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1362.0 Actor loss: -9.5075 Critic loss: +0.0051 World model loss: +0.7523 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4403 Episode_Reward/track_angular_velocity: +51.5690 Episode_Reward/upright: +26.1946 Episode_Reward/leg_posture: -2.5821 Episode_Reward/standstill: -5.8613 Episode_Reward/action_rate_l2: -0.5954 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.1047 Episode_Reward/rolling_leg_motion: -0.6371 Episode_Reward/obstacle_progress: +3.4232 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.3219 Curriculum/terrain_level_mean: +3.5181 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.4968 Episode_Termination/fell_over: 0.0318 Episode_Termination/course_exit: 0.4650 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 23850/50000 Total steps: 97,689,600 Steps per second: 12,467 Iteration time: 0.329s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1331.7 Actor loss: -9.5066 Critic loss: +0.0041 World model loss: +0.7005 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2738 Episode_Reward/track_angular_velocity: +50.3839 Episode_Reward/upright: +25.5912 Episode_Reward/leg_posture: -2.5064 Episode_Reward/standstill: -5.7528 Episode_Reward/action_rate_l2: -0.5810 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.0971 Episode_Reward/rolling_leg_motion: -0.6149 Episode_Reward/obstacle_progress: +3.3655 Episode_Reward/wheel_body_intrusion: -0.2474 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +3.5225 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4494 Episode_Termination/fell_over: 0.0443 Episode_Termination/course_exit: 0.4873 Episode_Termination/illegal_nonwheel_contact: 0.0443 ============================================================================== ============================================================================== Learning iteration 23900/50000 Total steps: 97,894,400 Steps per second: 12,473 Iteration time: 0.328s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1355.4 Actor loss: -9.5133 Critic loss: +0.0057 World model loss: +0.7168 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1465 Episode_Reward/track_angular_velocity: +51.2645 Episode_Reward/upright: +26.0240 Episode_Reward/leg_posture: -2.5850 Episode_Reward/standstill: -5.7778 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0937 Episode_Reward/rolling_leg_motion: -0.6278 Episode_Reward/obstacle_progress: +3.4079 Episode_Reward/wheel_body_intrusion: -0.2580 Episode_Reward/pair_body_intrusion: -0.3057 Curriculum/terrain_level_mean: +3.5251 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4938 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4568 Episode_Termination/illegal_nonwheel_contact: 0.0185 ============================================================================== ============================================================================== Learning iteration 23950/50000 Total steps: 98,099,200 Steps per second: 12,478 Iteration time: 0.328s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1398.5 Actor loss: -9.5077 Critic loss: +0.0043 World model loss: +0.6914 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6806 Episode_Reward/track_angular_velocity: +52.9069 Episode_Reward/upright: +26.8320 Episode_Reward/leg_posture: -2.7076 Episode_Reward/standstill: -5.8355 Episode_Reward/action_rate_l2: -0.6064 Episode_Reward/joint_torques_l2: -0.1529 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0913 Episode_Reward/rolling_leg_motion: -0.6395 Episode_Reward/obstacle_progress: +3.5522 Episode_Reward/wheel_body_intrusion: -0.2787 Episode_Reward/pair_body_intrusion: -0.3361 Curriculum/terrain_level_mean: +3.5298 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5070 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.4577 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== ============================================================================== Learning iteration 24000/50000 Total steps: 98,304,000 Steps per second: 12,483 Iteration time: 0.328s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0872 Mean episode length: 1358.0 Actor loss: -9.5228 Critic loss: +0.0027 World model loss: +0.6925 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0593 Episode_Reward/track_angular_velocity: +51.3349 Episode_Reward/upright: +26.0650 Episode_Reward/leg_posture: -2.5813 Episode_Reward/standstill: -5.7592 Episode_Reward/action_rate_l2: -0.6086 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.1045 Episode_Reward/rolling_leg_motion: -0.6437 Episode_Reward/obstacle_progress: +3.4259 Episode_Reward/wheel_body_intrusion: -0.2624 Episode_Reward/pair_body_intrusion: -0.3696 Curriculum/terrain_level_mean: +3.5317 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4830 Episode_Termination/fell_over: 0.0408 Episode_Termination/course_exit: 0.4626 Episode_Termination/illegal_nonwheel_contact: 0.0272 ============================================================================== ============================================================================== Learning iteration 24050/50000 Total steps: 98,508,800 Steps per second: 12,489 Iteration time: 0.328s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1352.5 Actor loss: -9.4885 Critic loss: +0.0058 World model loss: +0.6593 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8587 Episode_Reward/track_angular_velocity: +51.0961 Episode_Reward/upright: +25.9598 Episode_Reward/leg_posture: -2.5894 Episode_Reward/standstill: -5.5626 Episode_Reward/action_rate_l2: -0.6001 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0898 Episode_Reward/rolling_leg_motion: -0.6284 Episode_Reward/obstacle_progress: +3.4251 Episode_Reward/wheel_body_intrusion: -0.2645 Episode_Reward/pair_body_intrusion: -0.3783 Curriculum/terrain_level_mean: +3.5369 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4933 Episode_Termination/fell_over: 0.0133 Episode_Termination/course_exit: 0.4733 Episode_Termination/illegal_nonwheel_contact: 0.0200 ============================================================================== ============================================================================== Learning iteration 24100/50000 Total steps: 98,713,600 Steps per second: 12,494 Iteration time: 0.328s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1385.0 Actor loss: -9.5159 Critic loss: +0.0044 World model loss: +0.7059 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0722 Episode_Reward/track_angular_velocity: +52.2939 Episode_Reward/upright: +26.5694 Episode_Reward/leg_posture: -2.6386 Episode_Reward/standstill: -5.5970 Episode_Reward/action_rate_l2: -0.6270 Episode_Reward/joint_torques_l2: -0.1526 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.1102 Episode_Reward/rolling_leg_motion: -0.6753 Episode_Reward/obstacle_progress: +3.4697 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.3367 Curriculum/terrain_level_mean: +3.5403 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0203 Episode_Termination/course_exit: 0.4932 Episode_Termination/illegal_nonwheel_contact: 0.0405 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1937] RECOVERY reset 1 invalid environments; first IDs=[1388] ============================================================================== Learning iteration 24150/50000 Total steps: 98,918,400 Steps per second: 12,499 Iteration time: 0.328s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1409.6 Actor loss: -9.4740 Critic loss: +0.0065 World model loss: +0.7213 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0038 Episode_Reward/track_angular_velocity: +53.2062 Episode_Reward/upright: +27.0701 Episode_Reward/leg_posture: -2.7048 Episode_Reward/standstill: -5.6490 Episode_Reward/action_rate_l2: -0.6227 Episode_Reward/joint_torques_l2: -0.1545 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0193 Episode_Reward/unnecessary_wheel_lift: -0.1074 Episode_Reward/rolling_leg_motion: -0.6660 Episode_Reward/obstacle_progress: +3.5351 Episode_Reward/wheel_body_intrusion: -0.2557 Episode_Reward/pair_body_intrusion: -0.3114 Curriculum/terrain_level_mean: +3.5435 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.5411 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4110 Episode_Termination/illegal_nonwheel_contact: 0.0342 ============================================================================== ============================================================================== Learning iteration 24200/50000 Total steps: 99,123,200 Steps per second: 12,505 Iteration time: 0.328s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1412.3 Actor loss: -9.5013 Critic loss: +0.0039 World model loss: +0.7312 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1016 Episode_Reward/track_angular_velocity: +53.3033 Episode_Reward/upright: +27.1426 Episode_Reward/leg_posture: -2.7309 Episode_Reward/standstill: -5.7059 Episode_Reward/action_rate_l2: -0.6105 Episode_Reward/joint_torques_l2: -0.1567 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.1067 Episode_Reward/rolling_leg_motion: -0.6506 Episode_Reward/obstacle_progress: +3.5083 Episode_Reward/wheel_body_intrusion: -0.2619 Episode_Reward/pair_body_intrusion: -0.3330 Curriculum/terrain_level_mean: +3.5466 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4805 Episode_Termination/fell_over: 0.0195 Episode_Termination/course_exit: 0.4805 Episode_Termination/illegal_nonwheel_contact: 0.0325 ============================================================================== ============================================================================== Learning iteration 24250/50000 Total steps: 99,328,000 Steps per second: 12,510 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1331.9 Actor loss: -9.5033 Critic loss: +0.0044 World model loss: +0.7014 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2953 Episode_Reward/track_angular_velocity: +50.2477 Episode_Reward/upright: +25.5997 Episode_Reward/leg_posture: -2.5296 Episode_Reward/standstill: -5.6333 Episode_Reward/action_rate_l2: -0.5733 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0286 Episode_Reward/unnecessary_wheel_lift: -0.0965 Episode_Reward/rolling_leg_motion: -0.6356 Episode_Reward/obstacle_progress: +3.2869 Episode_Reward/wheel_body_intrusion: -0.2465 Episode_Reward/pair_body_intrusion: -0.3484 Curriculum/terrain_level_mean: +3.5493 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4731 Episode_Termination/fell_over: 0.0376 Episode_Termination/course_exit: 0.4677 Episode_Termination/illegal_nonwheel_contact: 0.0215 ============================================================================== ============================================================================== Learning iteration 24300/50000 Total steps: 99,532,800 Steps per second: 12,515 Iteration time: 0.327s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1437.8 Actor loss: -9.4676 Critic loss: +0.0056 World model loss: +0.6964 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.1858 Episode_Reward/track_angular_velocity: +54.2827 Episode_Reward/upright: +27.6202 Episode_Reward/leg_posture: -2.7489 Episode_Reward/standstill: -5.9878 Episode_Reward/action_rate_l2: -0.6053 Episode_Reward/joint_torques_l2: -0.1576 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0084 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.6570 Episode_Reward/obstacle_progress: +3.5763 Episode_Reward/wheel_body_intrusion: -0.2833 Episode_Reward/pair_body_intrusion: -0.3429 Curriculum/terrain_level_mean: +3.5530 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5591 Episode_Termination/fell_over: 0.0079 Episode_Termination/course_exit: 0.4173 Episode_Termination/illegal_nonwheel_contact: 0.0236 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_024300.mp4 ============================================================================== Learning iteration 24350/50000 Total steps: 99,737,600 Steps per second: 12,511 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0872 Mean episode length: 1281.4 Actor loss: -9.5081 Critic loss: +0.0047 World model loss: +0.6854 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.4884 Episode_Reward/track_angular_velocity: +48.3859 Episode_Reward/upright: +24.6305 Episode_Reward/leg_posture: -2.4590 Episode_Reward/standstill: -5.4872 Episode_Reward/action_rate_l2: -0.5520 Episode_Reward/joint_torques_l2: -0.1411 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0100 Episode_Reward/unnecessary_wheel_lift: -0.1041 Episode_Reward/rolling_leg_motion: -0.6132 Episode_Reward/obstacle_progress: +3.1845 Episode_Reward/wheel_body_intrusion: -0.2452 Episode_Reward/pair_body_intrusion: -0.3172 Curriculum/terrain_level_mean: +3.5591 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.3865 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5890 Episode_Termination/illegal_nonwheel_contact: 0.0245 ============================================================================== ============================================================================== Learning iteration 24400/50000 Total steps: 99,942,400 Steps per second: 12,516 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1350.7 Actor loss: -9.5101 Critic loss: +0.0030 World model loss: +0.7197 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0020 Episode_Reward/track_angular_velocity: +51.0575 Episode_Reward/upright: +25.9362 Episode_Reward/leg_posture: -2.6290 Episode_Reward/standstill: -5.8809 Episode_Reward/action_rate_l2: -0.5799 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.1096 Episode_Reward/rolling_leg_motion: -0.6316 Episode_Reward/obstacle_progress: +3.4208 Episode_Reward/wheel_body_intrusion: -0.2502 Episode_Reward/pair_body_intrusion: -0.3526 Curriculum/terrain_level_mean: +3.5618 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4583 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.5069 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== ============================================================================== Learning iteration 24450/50000 Total steps: 100,147,200 Steps per second: 12,521 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1298.0 Actor loss: -9.5052 Critic loss: +0.0042 World model loss: +0.7199 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0992 Episode_Reward/track_angular_velocity: +49.0075 Episode_Reward/upright: +24.9038 Episode_Reward/leg_posture: -2.5017 Episode_Reward/standstill: -5.6924 Episode_Reward/action_rate_l2: -0.5605 Episode_Reward/joint_torques_l2: -0.1420 Episode_Reward/dof_pos_limits: -0.0019 Episode_Reward/terminated: -0.0330 Episode_Reward/unnecessary_wheel_lift: -0.0919 Episode_Reward/rolling_leg_motion: -0.5860 Episode_Reward/obstacle_progress: +3.3422 Episode_Reward/wheel_body_intrusion: -0.2548 Episode_Reward/pair_body_intrusion: -0.3716 Curriculum/terrain_level_mean: +3.5632 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4507 Episode_Termination/fell_over: 0.0563 Episode_Termination/course_exit: 0.4577 Episode_Termination/illegal_nonwheel_contact: 0.0704 ============================================================================== ============================================================================== Learning iteration 24500/50000 Total steps: 100,352,000 Steps per second: 12,526 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1325.4 Actor loss: -9.5306 Critic loss: +0.0052 World model loss: +0.7134 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0547 Episode_Reward/track_angular_velocity: +50.0764 Episode_Reward/upright: +25.4766 Episode_Reward/leg_posture: -2.5789 Episode_Reward/standstill: -5.7671 Episode_Reward/action_rate_l2: -0.5682 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0337 Episode_Reward/unnecessary_wheel_lift: -0.0992 Episode_Reward/rolling_leg_motion: -0.6086 Episode_Reward/obstacle_progress: +3.3523 Episode_Reward/wheel_body_intrusion: -0.2507 Episode_Reward/pair_body_intrusion: -0.3673 Curriculum/terrain_level_mean: +3.5671 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4641 Episode_Termination/fell_over: 0.0719 Episode_Termination/course_exit: 0.4510 Episode_Termination/illegal_nonwheel_contact: 0.0327 ============================================================================== ============================================================================== Learning iteration 24550/50000 Total steps: 100,556,800 Steps per second: 12,531 Iteration time: 0.327s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1386.3 Actor loss: -9.4920 Critic loss: +0.0060 World model loss: +0.7004 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1865 Episode_Reward/track_angular_velocity: +52.4062 Episode_Reward/upright: +26.6165 Episode_Reward/leg_posture: -2.6673 Episode_Reward/standstill: -6.0991 Episode_Reward/action_rate_l2: -0.6040 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0784 Episode_Reward/rolling_leg_motion: -0.6289 Episode_Reward/obstacle_progress: +3.5616 Episode_Reward/wheel_body_intrusion: -0.2629 Episode_Reward/pair_body_intrusion: -0.3339 Curriculum/terrain_level_mean: +3.5688 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5118 Episode_Termination/fell_over: 0.0236 Episode_Termination/course_exit: 0.4409 Episode_Termination/illegal_nonwheel_contact: 0.0236 ============================================================================== ============================================================================== Learning iteration 24600/50000 Total steps: 100,761,600 Steps per second: 12,537 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1292.8 Actor loss: -9.4944 Critic loss: +0.0037 World model loss: +0.6940 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.6671 Episode_Reward/track_angular_velocity: +48.8369 Episode_Reward/upright: +24.8162 Episode_Reward/leg_posture: -2.5263 Episode_Reward/standstill: -5.8653 Episode_Reward/action_rate_l2: -0.5663 Episode_Reward/joint_torques_l2: -0.1433 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0303 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.5920 Episode_Reward/obstacle_progress: +3.2900 Episode_Reward/wheel_body_intrusion: -0.2265 Episode_Reward/pair_body_intrusion: -0.2573 Curriculum/terrain_level_mean: +3.5708 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4241 Episode_Termination/fell_over: 0.0570 Episode_Termination/course_exit: 0.5063 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== ============================================================================== Learning iteration 24650/50000 Total steps: 100,966,400 Steps per second: 12,542 Iteration time: 0.327s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1407.8 Actor loss: -9.4909 Critic loss: +0.0055 World model loss: +0.7093 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7295 Episode_Reward/track_angular_velocity: +53.1767 Episode_Reward/upright: +27.0279 Episode_Reward/leg_posture: -2.7367 Episode_Reward/standstill: -6.2594 Episode_Reward/action_rate_l2: -0.6265 Episode_Reward/joint_torques_l2: -0.1550 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0275 Episode_Reward/unnecessary_wheel_lift: -0.1030 Episode_Reward/rolling_leg_motion: -0.6645 Episode_Reward/obstacle_progress: +3.5215 Episode_Reward/wheel_body_intrusion: -0.2725 Episode_Reward/pair_body_intrusion: -0.3137 Curriculum/terrain_level_mean: +3.5742 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5460 Episode_Termination/fell_over: 0.0245 Episode_Termination/course_exit: 0.3926 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 24700/50000 Total steps: 101,171,200 Steps per second: 12,547 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1332.8 Actor loss: -9.5120 Critic loss: +0.0046 World model loss: +0.7062 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2704 Episode_Reward/track_angular_velocity: +50.3750 Episode_Reward/upright: +25.6048 Episode_Reward/leg_posture: -2.6026 Episode_Reward/standstill: -5.8560 Episode_Reward/action_rate_l2: -0.5805 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.0883 Episode_Reward/rolling_leg_motion: -0.6293 Episode_Reward/obstacle_progress: +3.3998 Episode_Reward/wheel_body_intrusion: -0.2644 Episode_Reward/pair_body_intrusion: -0.3089 Curriculum/terrain_level_mean: +3.5798 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4528 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.5094 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 24750/50000 Total steps: 101,376,000 Steps per second: 12,552 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0870 Mean episode length: 1329.5 Actor loss: -9.5097 Critic loss: +0.0029 World model loss: +0.6811 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0583 Episode_Reward/track_angular_velocity: +50.2429 Episode_Reward/upright: +25.5441 Episode_Reward/leg_posture: -2.6081 Episode_Reward/standstill: -5.8540 Episode_Reward/action_rate_l2: -0.5827 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0234 Episode_Reward/unnecessary_wheel_lift: -0.1098 Episode_Reward/rolling_leg_motion: -0.6211 Episode_Reward/obstacle_progress: +3.3590 Episode_Reward/wheel_body_intrusion: -0.2521 Episode_Reward/pair_body_intrusion: -0.3118 Curriculum/terrain_level_mean: +3.5833 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4307 Episode_Termination/fell_over: 0.0365 Episode_Termination/course_exit: 0.5036 Episode_Termination/illegal_nonwheel_contact: 0.0438 ============================================================================== ============================================================================== Learning iteration 24800/50000 Total steps: 101,580,800 Steps per second: 12,557 Iteration time: 0.326s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1349.8 Actor loss: -9.5087 Critic loss: +0.0040 World model loss: +0.7113 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8703 Episode_Reward/track_angular_velocity: +51.0313 Episode_Reward/upright: +25.9154 Episode_Reward/leg_posture: -2.6164 Episode_Reward/standstill: -6.0735 Episode_Reward/action_rate_l2: -0.5909 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0211 Episode_Reward/unnecessary_wheel_lift: -0.0909 Episode_Reward/rolling_leg_motion: -0.6146 Episode_Reward/obstacle_progress: +3.4774 Episode_Reward/wheel_body_intrusion: -0.2673 Episode_Reward/pair_body_intrusion: -0.3360 Curriculum/terrain_level_mean: +3.5840 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4500 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.4929 Episode_Termination/illegal_nonwheel_contact: 0.0357 ============================================================================== ============================================================================== Learning iteration 24850/50000 Total steps: 101,785,600 Steps per second: 12,562 Iteration time: 0.326s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1344.9 Actor loss: -9.5216 Critic loss: +0.0047 World model loss: +0.7023 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7177 Episode_Reward/track_angular_velocity: +50.8406 Episode_Reward/upright: +25.8198 Episode_Reward/leg_posture: -2.6318 Episode_Reward/standstill: -6.2618 Episode_Reward/action_rate_l2: -0.5885 Episode_Reward/joint_torques_l2: -0.1468 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.1000 Episode_Reward/rolling_leg_motion: -0.6202 Episode_Reward/obstacle_progress: +3.4327 Episode_Reward/wheel_body_intrusion: -0.2584 Episode_Reward/pair_body_intrusion: -0.3081 Curriculum/terrain_level_mean: +3.5864 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4348 Episode_Termination/fell_over: 0.0362 Episode_Termination/course_exit: 0.5145 Episode_Termination/illegal_nonwheel_contact: 0.0290 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2233] ============================================================================== Learning iteration 24900/50000 Total steps: 101,990,400 Steps per second: 12,568 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1424.7 Actor loss: -9.5214 Critic loss: +0.0038 World model loss: +0.6827 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.5858 Episode_Reward/track_angular_velocity: +53.8750 Episode_Reward/upright: +27.3554 Episode_Reward/leg_posture: -2.7761 Episode_Reward/standstill: -6.6151 Episode_Reward/action_rate_l2: -0.6138 Episode_Reward/joint_torques_l2: -0.1575 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0223 Episode_Reward/unnecessary_wheel_lift: -0.0992 Episode_Reward/rolling_leg_motion: -0.6494 Episode_Reward/obstacle_progress: +3.6051 Episode_Reward/wheel_body_intrusion: -0.2689 Episode_Reward/pair_body_intrusion: -0.3042 Curriculum/terrain_level_mean: +3.5891 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5540 Episode_Termination/fell_over: 0.0288 Episode_Termination/course_exit: 0.3885 Episode_Termination/illegal_nonwheel_contact: 0.0504 ============================================================================== ============================================================================== Learning iteration 24950/50000 Total steps: 102,195,200 Steps per second: 12,573 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1403.5 Actor loss: -9.5059 Critic loss: +0.0045 World model loss: +0.6940 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6567 Episode_Reward/track_angular_velocity: +53.0872 Episode_Reward/upright: +26.9558 Episode_Reward/leg_posture: -2.7551 Episode_Reward/standstill: -6.6301 Episode_Reward/action_rate_l2: -0.6114 Episode_Reward/joint_torques_l2: -0.1556 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0082 Episode_Reward/unnecessary_wheel_lift: -0.1118 Episode_Reward/rolling_leg_motion: -0.6488 Episode_Reward/obstacle_progress: +3.5253 Episode_Reward/wheel_body_intrusion: -0.2571 Episode_Reward/pair_body_intrusion: -0.2777 Curriculum/terrain_level_mean: +3.5950 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.5423 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.4366 Episode_Termination/illegal_nonwheel_contact: 0.0141 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1950] ============================================================================== Learning iteration 25000/50000 Total steps: 102,400,000 Steps per second: 12,577 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0865 Mean episode length: 1353.0 Actor loss: -9.4888 Critic loss: +0.0038 World model loss: +0.6713 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9245 Episode_Reward/track_angular_velocity: +51.0942 Episode_Reward/upright: +25.9474 Episode_Reward/leg_posture: -2.6930 Episode_Reward/standstill: -6.6229 Episode_Reward/action_rate_l2: -0.6023 Episode_Reward/joint_torques_l2: -0.1497 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0317 Episode_Reward/unnecessary_wheel_lift: -0.0938 Episode_Reward/rolling_leg_motion: -0.6242 Episode_Reward/obstacle_progress: +3.4218 Episode_Reward/wheel_body_intrusion: -0.2673 Episode_Reward/pair_body_intrusion: -0.3249 Curriculum/terrain_level_mean: +3.5974 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4610 Episode_Termination/fell_over: 0.0455 Episode_Termination/course_exit: 0.4610 Episode_Termination/illegal_nonwheel_contact: 0.0519 ============================================================================== ============================================================================== Learning iteration 25050/50000 Total steps: 102,604,800 Steps per second: 12,582 Iteration time: 0.326s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0868 Mean episode length: 1295.2 Actor loss: -9.4912 Critic loss: +0.0058 World model loss: +0.6641 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8036 Episode_Reward/track_angular_velocity: +48.9165 Episode_Reward/upright: +24.8565 Episode_Reward/leg_posture: -2.5487 Episode_Reward/standstill: -5.8360 Episode_Reward/action_rate_l2: -0.5681 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0354 Episode_Reward/unnecessary_wheel_lift: -0.0830 Episode_Reward/rolling_leg_motion: -0.6042 Episode_Reward/obstacle_progress: +3.2252 Episode_Reward/wheel_body_intrusion: -0.2545 Episode_Reward/pair_body_intrusion: -0.3356 Curriculum/terrain_level_mean: +3.5986 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4315 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.4726 Episode_Termination/illegal_nonwheel_contact: 0.0753 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2598] ============================================================================== Learning iteration 25100/50000 Total steps: 102,809,600 Steps per second: 12,587 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0869 Mean episode length: 1327.3 Actor loss: -9.4774 Critic loss: +0.0110 World model loss: +0.6755 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9167 Episode_Reward/track_angular_velocity: +50.1103 Episode_Reward/upright: +25.4668 Episode_Reward/leg_posture: -2.6137 Episode_Reward/standstill: -5.8954 Episode_Reward/action_rate_l2: -0.5719 Episode_Reward/joint_torques_l2: -0.1470 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.0961 Episode_Reward/rolling_leg_motion: -0.6077 Episode_Reward/obstacle_progress: +3.3225 Episode_Reward/wheel_body_intrusion: -0.2515 Episode_Reward/pair_body_intrusion: -0.3291 Curriculum/terrain_level_mean: +3.6023 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4268 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.5223 Episode_Termination/illegal_nonwheel_contact: 0.0382 ============================================================================== ============================================================================== Learning iteration 25150/50000 Total steps: 103,014,400 Steps per second: 12,592 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1331.5 Actor loss: -9.5128 Critic loss: +0.0036 World model loss: +0.7525 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0691 Episode_Reward/track_angular_velocity: +50.2398 Episode_Reward/upright: +25.5431 Episode_Reward/leg_posture: -2.6497 Episode_Reward/standstill: -5.8896 Episode_Reward/action_rate_l2: -0.5829 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.1073 Episode_Reward/rolling_leg_motion: -0.6150 Episode_Reward/obstacle_progress: +3.3395 Episode_Reward/wheel_body_intrusion: -0.2644 Episode_Reward/pair_body_intrusion: -0.3561 Curriculum/terrain_level_mean: +3.6035 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4431 Episode_Termination/fell_over: 0.0299 Episode_Termination/course_exit: 0.4970 Episode_Termination/illegal_nonwheel_contact: 0.0359 ============================================================================== ============================================================================== Learning iteration 25200/50000 Total steps: 103,219,200 Steps per second: 12,597 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1326.4 Actor loss: -9.5040 Critic loss: +0.0061 World model loss: +0.6473 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8494 Episode_Reward/track_angular_velocity: +50.0874 Episode_Reward/upright: +25.4630 Episode_Reward/leg_posture: -2.6149 Episode_Reward/standstill: -6.0446 Episode_Reward/action_rate_l2: -0.5858 Episode_Reward/joint_torques_l2: -0.1471 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.1059 Episode_Reward/rolling_leg_motion: -0.6350 Episode_Reward/obstacle_progress: +3.3111 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.3126 Curriculum/terrain_level_mean: +3.6069 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4902 Episode_Termination/fell_over: 0.0392 Episode_Termination/course_exit: 0.4641 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 25250/50000 Total steps: 103,424,000 Steps per second: 12,602 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1393.8 Actor loss: -9.4570 Critic loss: +0.0094 World model loss: +0.6778 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3432 Episode_Reward/track_angular_velocity: +52.6456 Episode_Reward/upright: +26.7568 Episode_Reward/leg_posture: -2.6875 Episode_Reward/standstill: -6.3784 Episode_Reward/action_rate_l2: -0.6136 Episode_Reward/joint_torques_l2: -0.1542 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0993 Episode_Reward/rolling_leg_motion: -0.6481 Episode_Reward/obstacle_progress: +3.5128 Episode_Reward/wheel_body_intrusion: -0.2613 Episode_Reward/pair_body_intrusion: -0.3074 Curriculum/terrain_level_mean: +3.6094 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4875 Episode_Termination/fell_over: 0.0187 Episode_Termination/course_exit: 0.4875 Episode_Termination/illegal_nonwheel_contact: 0.0125 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1647] ============================================================================== Learning iteration 25300/50000 Total steps: 103,628,800 Steps per second: 12,607 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1346.0 Actor loss: -9.4999 Critic loss: +0.0050 World model loss: +0.7107 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5483 Episode_Reward/track_angular_velocity: +50.9271 Episode_Reward/upright: +25.8429 Episode_Reward/leg_posture: -2.6140 Episode_Reward/standstill: -5.8997 Episode_Reward/action_rate_l2: -0.5850 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0946 Episode_Reward/rolling_leg_motion: -0.6405 Episode_Reward/obstacle_progress: +3.2643 Episode_Reward/wheel_body_intrusion: -0.2494 Episode_Reward/pair_body_intrusion: -0.3216 Curriculum/terrain_level_mean: +3.6152 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4296 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.5423 Episode_Termination/illegal_nonwheel_contact: 0.0141 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1163] ============================================================================== Learning iteration 25350/50000 Total steps: 103,833,600 Steps per second: 12,612 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1331.0 Actor loss: -9.4734 Critic loss: +0.0041 World model loss: +0.6361 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0988 Episode_Reward/track_angular_velocity: +50.3609 Episode_Reward/upright: +25.5377 Episode_Reward/leg_posture: -2.5757 Episode_Reward/standstill: -5.7681 Episode_Reward/action_rate_l2: -0.5752 Episode_Reward/joint_torques_l2: -0.1468 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0312 Episode_Reward/unnecessary_wheel_lift: -0.0861 Episode_Reward/rolling_leg_motion: -0.6222 Episode_Reward/obstacle_progress: +3.2735 Episode_Reward/wheel_body_intrusion: -0.2797 Episode_Reward/pair_body_intrusion: -0.2633 Curriculum/terrain_level_mean: +3.6179 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4452 Episode_Termination/fell_over: 0.0516 Episode_Termination/course_exit: 0.4774 Episode_Termination/illegal_nonwheel_contact: 0.0387 ============================================================================== ============================================================================== Learning iteration 25400/50000 Total steps: 104,038,400 Steps per second: 12,617 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1352.1 Actor loss: -9.4729 Critic loss: +0.0041 World model loss: +0.6759 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8444 Episode_Reward/track_angular_velocity: +51.1049 Episode_Reward/upright: +25.9595 Episode_Reward/leg_posture: -2.6218 Episode_Reward/standstill: -6.1845 Episode_Reward/action_rate_l2: -0.5727 Episode_Reward/joint_torques_l2: -0.1489 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0770 Episode_Reward/rolling_leg_motion: -0.6152 Episode_Reward/obstacle_progress: +3.3420 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.2993 Curriculum/terrain_level_mean: +3.6208 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4741 Episode_Termination/fell_over: 0.0259 Episode_Termination/course_exit: 0.4828 Episode_Termination/illegal_nonwheel_contact: 0.0431 ============================================================================== ============================================================================== Learning iteration 25450/50000 Total steps: 104,243,200 Steps per second: 12,622 Iteration time: 0.325s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1396.8 Actor loss: -9.4785 Critic loss: +0.0060 World model loss: +0.6699 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5384 Episode_Reward/track_angular_velocity: +52.7932 Episode_Reward/upright: +26.7910 Episode_Reward/leg_posture: -2.7194 Episode_Reward/standstill: -5.9911 Episode_Reward/action_rate_l2: -0.5899 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0191 Episode_Reward/unnecessary_wheel_lift: -0.0931 Episode_Reward/rolling_leg_motion: -0.6473 Episode_Reward/obstacle_progress: +3.5470 Episode_Reward/wheel_body_intrusion: -0.2797 Episode_Reward/pair_body_intrusion: -0.3111 Curriculum/terrain_level_mean: +3.6240 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4929 Episode_Termination/fell_over: 0.0429 Episode_Termination/course_exit: 0.4571 Episode_Termination/illegal_nonwheel_contact: 0.0214 ============================================================================== ============================================================================== Learning iteration 25500/50000 Total steps: 104,448,000 Steps per second: 12,626 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1370.5 Actor loss: -9.5010 Critic loss: +0.0036 World model loss: +0.7031 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4804 Episode_Reward/track_angular_velocity: +51.7642 Episode_Reward/upright: +26.2660 Episode_Reward/leg_posture: -2.6654 Episode_Reward/standstill: -6.1605 Episode_Reward/action_rate_l2: -0.6022 Episode_Reward/joint_torques_l2: -0.1508 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0976 Episode_Reward/rolling_leg_motion: -0.6416 Episode_Reward/obstacle_progress: +3.5173 Episode_Reward/wheel_body_intrusion: -0.2659 Episode_Reward/pair_body_intrusion: -0.3300 Curriculum/terrain_level_mean: +3.6299 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4793 Episode_Termination/fell_over: 0.0059 Episode_Termination/course_exit: 0.4911 Episode_Termination/illegal_nonwheel_contact: 0.0296 ============================================================================== ============================================================================== Learning iteration 25550/50000 Total steps: 104,652,800 Steps per second: 12,631 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0868 Mean episode length: 1358.7 Actor loss: -9.5206 Critic loss: +0.0031 World model loss: +0.6447 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0567 Episode_Reward/track_angular_velocity: +51.3937 Episode_Reward/upright: +26.0454 Episode_Reward/leg_posture: -2.6658 Episode_Reward/standstill: -6.1533 Episode_Reward/action_rate_l2: -0.5869 Episode_Reward/joint_torques_l2: -0.1489 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0239 Episode_Reward/unnecessary_wheel_lift: -0.0977 Episode_Reward/rolling_leg_motion: -0.6294 Episode_Reward/obstacle_progress: +3.3846 Episode_Reward/wheel_body_intrusion: -0.2533 Episode_Reward/pair_body_intrusion: -0.3219 Curriculum/terrain_level_mean: +3.6323 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.5169 Episode_Termination/fell_over: 0.0337 Episode_Termination/course_exit: 0.4326 Episode_Termination/illegal_nonwheel_contact: 0.0169 ============================================================================== ============================================================================== Learning iteration 25600/50000 Total steps: 104,857,600 Steps per second: 12,636 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1386.7 Actor loss: -9.5015 Critic loss: +0.0036 World model loss: +0.7006 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1799 Episode_Reward/track_angular_velocity: +52.4293 Episode_Reward/upright: +26.6234 Episode_Reward/leg_posture: -2.7114 Episode_Reward/standstill: -6.1169 Episode_Reward/action_rate_l2: -0.5835 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0163 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.6325 Episode_Reward/obstacle_progress: +3.4703 Episode_Reward/wheel_body_intrusion: -0.2663 Episode_Reward/pair_body_intrusion: -0.3456 Curriculum/terrain_level_mean: +3.6360 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.5347 Episode_Termination/fell_over: 0.0347 Episode_Termination/course_exit: 0.4236 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== ============================================================================== Learning iteration 25650/50000 Total steps: 105,062,400 Steps per second: 12,640 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1459.1 Actor loss: -9.4621 Critic loss: +0.0069 World model loss: +0.6681 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.6970 Episode_Reward/track_angular_velocity: +55.1812 Episode_Reward/upright: +27.9900 Episode_Reward/leg_posture: -2.8627 Episode_Reward/standstill: -6.7209 Episode_Reward/action_rate_l2: -0.6350 Episode_Reward/joint_torques_l2: -0.1598 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0171 Episode_Reward/unnecessary_wheel_lift: -0.0917 Episode_Reward/rolling_leg_motion: -0.6716 Episode_Reward/obstacle_progress: +3.6542 Episode_Reward/wheel_body_intrusion: -0.3006 Episode_Reward/pair_body_intrusion: -0.3862 Curriculum/terrain_level_mean: +3.6384 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5490 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.4118 Episode_Termination/illegal_nonwheel_contact: 0.0261 ============================================================================== ============================================================================== Learning iteration 25700/50000 Total steps: 105,267,200 Steps per second: 12,645 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1383.5 Actor loss: -9.4778 Critic loss: +0.0042 World model loss: +0.6582 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0038 Episode_Reward/track_angular_velocity: +52.2951 Episode_Reward/upright: +26.5345 Episode_Reward/leg_posture: -2.7208 Episode_Reward/standstill: -6.2264 Episode_Reward/action_rate_l2: -0.5788 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.6368 Episode_Reward/obstacle_progress: +3.4691 Episode_Reward/wheel_body_intrusion: -0.2845 Episode_Reward/pair_body_intrusion: -0.3654 Curriculum/terrain_level_mean: +3.6418 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.5221 Episode_Termination/fell_over: 0.0074 Episode_Termination/course_exit: 0.4485 Episode_Termination/illegal_nonwheel_contact: 0.0221 ============================================================================== ============================================================================== Learning iteration 25750/50000 Total steps: 105,472,000 Steps per second: 12,647 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1371.7 Actor loss: -9.4911 Critic loss: +0.0036 World model loss: +0.6484 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5665 Episode_Reward/track_angular_velocity: +51.8726 Episode_Reward/upright: +26.2758 Episode_Reward/leg_posture: -2.6613 Episode_Reward/standstill: -6.3193 Episode_Reward/action_rate_l2: -0.5735 Episode_Reward/joint_torques_l2: -0.1507 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.1001 Episode_Reward/rolling_leg_motion: -0.6188 Episode_Reward/obstacle_progress: +3.5497 Episode_Reward/wheel_body_intrusion: -0.2844 Episode_Reward/pair_body_intrusion: -0.3590 Curriculum/terrain_level_mean: +3.6455 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5063 Episode_Termination/fell_over: 0.0190 Episode_Termination/course_exit: 0.4494 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2369] ============================================================================== Learning iteration 25800/50000 Total steps: 105,676,800 Steps per second: 12,652 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1390.6 Actor loss: -9.4983 Critic loss: +0.0073 World model loss: +0.6725 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2045 Episode_Reward/track_angular_velocity: +52.6045 Episode_Reward/upright: +26.6663 Episode_Reward/leg_posture: -2.6711 Episode_Reward/standstill: -5.9186 Episode_Reward/action_rate_l2: -0.5913 Episode_Reward/joint_torques_l2: -0.1523 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0081 Episode_Reward/unnecessary_wheel_lift: -0.0835 Episode_Reward/rolling_leg_motion: -0.6168 Episode_Reward/obstacle_progress: +3.6205 Episode_Reward/wheel_body_intrusion: -0.2925 Episode_Reward/pair_body_intrusion: -0.3521 Curriculum/terrain_level_mean: +3.6499 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4872 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.4936 Episode_Termination/illegal_nonwheel_contact: 0.0064 ============================================================================== ============================================================================== Learning iteration 25850/50000 Total steps: 105,881,600 Steps per second: 12,656 Iteration time: 0.324s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1371.3 Actor loss: -9.5306 Critic loss: +0.0035 World model loss: +0.6697 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5011 Episode_Reward/track_angular_velocity: +51.8350 Episode_Reward/upright: +26.2910 Episode_Reward/leg_posture: -2.5912 Episode_Reward/standstill: -5.8632 Episode_Reward/action_rate_l2: -0.5791 Episode_Reward/joint_torques_l2: -0.1513 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0773 Episode_Reward/rolling_leg_motion: -0.6066 Episode_Reward/obstacle_progress: +3.6578 Episode_Reward/wheel_body_intrusion: -0.2790 Episode_Reward/pair_body_intrusion: -0.3386 Curriculum/terrain_level_mean: +3.6499 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4884 Episode_Termination/fell_over: 0.0078 Episode_Termination/course_exit: 0.4574 Episode_Termination/illegal_nonwheel_contact: 0.0465 ============================================================================== ============================================================================== Learning iteration 25900/50000 Total steps: 106,086,400 Steps per second: 12,661 Iteration time: 0.324s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1415.1 Actor loss: -9.5073 Critic loss: +0.0034 World model loss: +0.6723 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1128 Episode_Reward/track_angular_velocity: +53.4506 Episode_Reward/upright: +27.1444 Episode_Reward/leg_posture: -2.6496 Episode_Reward/standstill: -6.0852 Episode_Reward/action_rate_l2: -0.6122 Episode_Reward/joint_torques_l2: -0.1564 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0193 Episode_Reward/unnecessary_wheel_lift: -0.0825 Episode_Reward/rolling_leg_motion: -0.6299 Episode_Reward/obstacle_progress: +3.8534 Episode_Reward/wheel_body_intrusion: -0.2697 Episode_Reward/pair_body_intrusion: -0.3934 Curriculum/terrain_level_mean: +3.6519 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5299 Episode_Termination/fell_over: 0.0224 Episode_Termination/course_exit: 0.4179 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2044] ============================================================================== Learning iteration 25950/50000 Total steps: 106,291,200 Steps per second: 12,666 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1380.6 Actor loss: -9.4802 Critic loss: +0.0052 World model loss: +0.6615 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7177 Episode_Reward/track_angular_velocity: +52.1051 Episode_Reward/upright: +26.4809 Episode_Reward/leg_posture: -2.6303 Episode_Reward/standstill: -5.8843 Episode_Reward/action_rate_l2: -0.5804 Episode_Reward/joint_torques_l2: -0.1552 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.6207 Episode_Reward/obstacle_progress: +3.5838 Episode_Reward/wheel_body_intrusion: -0.3093 Episode_Reward/pair_body_intrusion: -0.3847 Curriculum/terrain_level_mean: +3.6555 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.5217 Episode_Termination/fell_over: 0.0217 Episode_Termination/course_exit: 0.4275 Episode_Termination/illegal_nonwheel_contact: 0.0290 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2323] ============================================================================== Learning iteration 26000/50000 Total steps: 106,496,000 Steps per second: 12,671 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1379.6 Actor loss: -9.5030 Critic loss: +0.0047 World model loss: +0.6879 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7079 Episode_Reward/track_angular_velocity: +52.0348 Episode_Reward/upright: +26.4853 Episode_Reward/leg_posture: -2.6435 Episode_Reward/standstill: -6.1722 Episode_Reward/action_rate_l2: -0.5945 Episode_Reward/joint_torques_l2: -0.1545 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0323 Episode_Reward/unnecessary_wheel_lift: -0.1041 Episode_Reward/rolling_leg_motion: -0.6397 Episode_Reward/obstacle_progress: +3.5550 Episode_Reward/wheel_body_intrusion: -0.2971 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +3.6580 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.5351 Episode_Termination/fell_over: 0.0486 Episode_Termination/course_exit: 0.4000 Episode_Termination/illegal_nonwheel_contact: 0.0324 ============================================================================== ============================================================================== Learning iteration 26050/50000 Total steps: 106,700,800 Steps per second: 12,675 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1350.6 Actor loss: -9.5063 Critic loss: +0.0047 World model loss: +0.6492 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7889 Episode_Reward/track_angular_velocity: +51.0392 Episode_Reward/upright: +25.9378 Episode_Reward/leg_posture: -2.5722 Episode_Reward/standstill: -5.9605 Episode_Reward/action_rate_l2: -0.5831 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0317 Episode_Reward/unnecessary_wheel_lift: -0.0934 Episode_Reward/rolling_leg_motion: -0.6194 Episode_Reward/obstacle_progress: +3.4448 Episode_Reward/wheel_body_intrusion: -0.2467 Episode_Reward/pair_body_intrusion: -0.2799 Curriculum/terrain_level_mean: +3.6602 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4641 Episode_Termination/fell_over: 0.0523 Episode_Termination/course_exit: 0.4575 Episode_Termination/illegal_nonwheel_contact: 0.0588 ============================================================================== ============================================================================== Learning iteration 26100/50000 Total steps: 106,905,600 Steps per second: 12,680 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1362.9 Actor loss: -9.4979 Critic loss: +0.0046 World model loss: +0.6704 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3885 Episode_Reward/track_angular_velocity: +51.5023 Episode_Reward/upright: +26.1964 Episode_Reward/leg_posture: -2.5896 Episode_Reward/standstill: -6.0782 Episode_Reward/action_rate_l2: -0.6024 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.0731 Episode_Reward/rolling_leg_motion: -0.6136 Episode_Reward/obstacle_progress: +3.5332 Episode_Reward/wheel_body_intrusion: -0.2464 Episode_Reward/pair_body_intrusion: -0.2758 Curriculum/terrain_level_mean: +3.6633 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4810 Episode_Termination/fell_over: 0.0253 Episode_Termination/course_exit: 0.4684 Episode_Termination/illegal_nonwheel_contact: 0.0316 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1233] ============================================================================== Learning iteration 26150/50000 Total steps: 107,110,400 Steps per second: 12,685 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1360.2 Actor loss: -9.4855 Critic loss: +0.0057 World model loss: +0.6859 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3371 Episode_Reward/track_angular_velocity: +51.4305 Episode_Reward/upright: +26.1613 Episode_Reward/leg_posture: -2.5884 Episode_Reward/standstill: -6.0963 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.0879 Episode_Reward/rolling_leg_motion: -0.6238 Episode_Reward/obstacle_progress: +3.4596 Episode_Reward/wheel_body_intrusion: -0.2671 Episode_Reward/pair_body_intrusion: -0.2650 Curriculum/terrain_level_mean: +3.6658 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4514 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.5139 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== ============================================================================== Learning iteration 26200/50000 Total steps: 107,315,200 Steps per second: 12,689 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1372.9 Actor loss: -9.4777 Critic loss: +0.0098 World model loss: +0.6652 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6902 Episode_Reward/track_angular_velocity: +51.8560 Episode_Reward/upright: +26.3663 Episode_Reward/leg_posture: -2.6225 Episode_Reward/standstill: -6.0264 Episode_Reward/action_rate_l2: -0.5931 Episode_Reward/joint_torques_l2: -0.1534 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0268 Episode_Reward/unnecessary_wheel_lift: -0.0922 Episode_Reward/rolling_leg_motion: -0.6415 Episode_Reward/obstacle_progress: +3.5197 Episode_Reward/wheel_body_intrusion: -0.2610 Episode_Reward/pair_body_intrusion: -0.3255 Curriculum/terrain_level_mean: +3.6692 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4653 Episode_Termination/fell_over: 0.0486 Episode_Termination/course_exit: 0.4653 Episode_Termination/illegal_nonwheel_contact: 0.0417 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2218] ============================================================================== Learning iteration 26250/50000 Total steps: 107,520,000 Steps per second: 12,694 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1366.3 Actor loss: -9.5038 Critic loss: +0.0081 World model loss: +0.6667 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5464 Episode_Reward/track_angular_velocity: +51.6374 Episode_Reward/upright: +26.2556 Episode_Reward/leg_posture: -2.5705 Episode_Reward/standstill: -6.0426 Episode_Reward/action_rate_l2: -0.5841 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0792 Episode_Reward/rolling_leg_motion: -0.6240 Episode_Reward/obstacle_progress: +3.5195 Episode_Reward/wheel_body_intrusion: -0.2575 Episode_Reward/pair_body_intrusion: -0.2925 Curriculum/terrain_level_mean: +3.6726 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4641 Episode_Termination/fell_over: 0.0196 Episode_Termination/course_exit: 0.5098 Episode_Termination/illegal_nonwheel_contact: 0.0196 ============================================================================== ============================================================================== Learning iteration 26300/50000 Total steps: 107,724,800 Steps per second: 12,699 Iteration time: 0.323s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1294.4 Actor loss: -9.5163 Critic loss: +0.0033 World model loss: +0.6484 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0620 Episode_Reward/track_angular_velocity: +48.9875 Episode_Reward/upright: +24.8908 Episode_Reward/leg_posture: -2.4867 Episode_Reward/standstill: -5.7592 Episode_Reward/action_rate_l2: -0.5483 Episode_Reward/joint_torques_l2: -0.1415 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0202 Episode_Reward/unnecessary_wheel_lift: -0.0829 Episode_Reward/rolling_leg_motion: -0.6007 Episode_Reward/obstacle_progress: +3.3448 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.3100 Curriculum/terrain_level_mean: +3.6733 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4306 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.5139 Episode_Termination/illegal_nonwheel_contact: 0.0486 ============================================================================== ============================================================================== Learning iteration 26350/50000 Total steps: 107,929,600 Steps per second: 12,703 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0873 Mean episode length: 1390.4 Actor loss: -9.4869 Critic loss: +0.0061 World model loss: +0.6590 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5097 Episode_Reward/track_angular_velocity: +52.6043 Episode_Reward/upright: +26.7337 Episode_Reward/leg_posture: -2.6836 Episode_Reward/standstill: -6.2752 Episode_Reward/action_rate_l2: -0.5984 Episode_Reward/joint_torques_l2: -0.1526 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.0869 Episode_Reward/rolling_leg_motion: -0.6379 Episode_Reward/obstacle_progress: +3.6117 Episode_Reward/wheel_body_intrusion: -0.2914 Episode_Reward/pair_body_intrusion: -0.3046 Curriculum/terrain_level_mean: +3.6746 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5321 Episode_Termination/fell_over: 0.0385 Episode_Termination/course_exit: 0.4167 Episode_Termination/illegal_nonwheel_contact: 0.0128 ============================================================================== ============================================================================== Learning iteration 26400/50000 Total steps: 108,134,400 Steps per second: 12,707 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1411.9 Actor loss: -9.5033 Critic loss: +0.0056 World model loss: +0.6489 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1741 Episode_Reward/track_angular_velocity: +53.4427 Episode_Reward/upright: +27.1281 Episode_Reward/leg_posture: -2.7071 Episode_Reward/standstill: -6.7598 Episode_Reward/action_rate_l2: -0.6141 Episode_Reward/joint_torques_l2: -0.1545 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1064 Episode_Reward/rolling_leg_motion: -0.6468 Episode_Reward/obstacle_progress: +3.6768 Episode_Reward/wheel_body_intrusion: -0.2973 Episode_Reward/pair_body_intrusion: -0.3310 Curriculum/terrain_level_mean: +3.6760 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.5443 Episode_Termination/fell_over: 0.0316 Episode_Termination/course_exit: 0.4051 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1474] ============================================================================== Learning iteration 26450/50000 Total steps: 108,339,200 Steps per second: 12,712 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1407.1 Actor loss: -9.5045 Critic loss: +0.0047 World model loss: +0.6445 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9458 Episode_Reward/track_angular_velocity: +53.2579 Episode_Reward/upright: +27.0767 Episode_Reward/leg_posture: -2.6455 Episode_Reward/standstill: -6.1471 Episode_Reward/action_rate_l2: -0.5990 Episode_Reward/joint_torques_l2: -0.1532 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0302 Episode_Reward/unnecessary_wheel_lift: -0.0870 Episode_Reward/rolling_leg_motion: -0.6410 Episode_Reward/obstacle_progress: +3.5897 Episode_Reward/wheel_body_intrusion: -0.2807 Episode_Reward/pair_body_intrusion: -0.3219 Curriculum/terrain_level_mean: +3.6780 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.5101 Episode_Termination/fell_over: 0.0470 Episode_Termination/course_exit: 0.4161 Episode_Termination/illegal_nonwheel_contact: 0.0537 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1778] ============================================================================== Learning iteration 26500/50000 Total steps: 108,544,000 Steps per second: 12,716 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1355.6 Actor loss: -9.5030 Critic loss: +0.0038 World model loss: +0.6600 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1961 Episode_Reward/track_angular_velocity: +51.3098 Episode_Reward/upright: +26.1238 Episode_Reward/leg_posture: -2.5593 Episode_Reward/standstill: -5.7390 Episode_Reward/action_rate_l2: -0.5706 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0725 Episode_Reward/rolling_leg_motion: -0.5889 Episode_Reward/obstacle_progress: +3.4783 Episode_Reward/wheel_body_intrusion: -0.2777 Episode_Reward/pair_body_intrusion: -0.3280 Curriculum/terrain_level_mean: +3.6807 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4748 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4676 Episode_Termination/illegal_nonwheel_contact: 0.0360 ============================================================================== ============================================================================== Learning iteration 26550/50000 Total steps: 108,748,800 Steps per second: 12,721 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1340.2 Actor loss: -9.5492 Critic loss: +0.0057 World model loss: +0.6349 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5361 Episode_Reward/track_angular_velocity: +50.6945 Episode_Reward/upright: +25.8006 Episode_Reward/leg_posture: -2.5834 Episode_Reward/standstill: -5.9663 Episode_Reward/action_rate_l2: -0.5658 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0183 Episode_Reward/unnecessary_wheel_lift: -0.0776 Episode_Reward/rolling_leg_motion: -0.5973 Episode_Reward/obstacle_progress: +3.4320 Episode_Reward/wheel_body_intrusion: -0.2871 Episode_Reward/pair_body_intrusion: -0.3599 Curriculum/terrain_level_mean: +3.6843 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4067 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.5467 Episode_Termination/illegal_nonwheel_contact: 0.0467 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1208] ============================================================================== Learning iteration 26600/50000 Total steps: 108,953,600 Steps per second: 12,726 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1325.1 Actor loss: -9.4936 Critic loss: +0.0064 World model loss: +0.6322 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9602 Episode_Reward/track_angular_velocity: +50.0416 Episode_Reward/upright: +25.4947 Episode_Reward/leg_posture: -2.5379 Episode_Reward/standstill: -5.7308 Episode_Reward/action_rate_l2: -0.5658 Episode_Reward/joint_torques_l2: -0.1449 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0855 Episode_Reward/rolling_leg_motion: -0.5962 Episode_Reward/obstacle_progress: +3.4290 Episode_Reward/wheel_body_intrusion: -0.2939 Episode_Reward/pair_body_intrusion: -0.3190 Curriculum/terrain_level_mean: +3.6873 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4380 Episode_Termination/fell_over: 0.0219 Episode_Termination/course_exit: 0.5182 Episode_Termination/illegal_nonwheel_contact: 0.0292 ============================================================================== ============================================================================== Learning iteration 26650/50000 Total steps: 109,158,400 Steps per second: 12,730 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1339.5 Actor loss: -9.5062 Critic loss: +0.0046 World model loss: +0.6318 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3251 Episode_Reward/track_angular_velocity: +50.6020 Episode_Reward/upright: +25.7507 Episode_Reward/leg_posture: -2.5421 Episode_Reward/standstill: -5.8126 Episode_Reward/action_rate_l2: -0.5824 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0881 Episode_Reward/rolling_leg_motion: -0.5990 Episode_Reward/obstacle_progress: +3.4540 Episode_Reward/wheel_body_intrusion: -0.2790 Episode_Reward/pair_body_intrusion: -0.3251 Curriculum/terrain_level_mean: +3.6929 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4483 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.5241 Episode_Termination/illegal_nonwheel_contact: 0.0207 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1396] ============================================================================== Learning iteration 26700/50000 Total steps: 109,363,200 Steps per second: 12,735 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1349.4 Actor loss: -9.5126 Critic loss: +0.0039 World model loss: +0.6082 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6379 Episode_Reward/track_angular_velocity: +50.9421 Episode_Reward/upright: +25.9099 Episode_Reward/leg_posture: -2.5500 Episode_Reward/standstill: -5.8576 Episode_Reward/action_rate_l2: -0.5877 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0888 Episode_Reward/rolling_leg_motion: -0.6139 Episode_Reward/obstacle_progress: +3.4564 Episode_Reward/wheel_body_intrusion: -0.2742 Episode_Reward/pair_body_intrusion: -0.3392 Curriculum/terrain_level_mean: +3.6958 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4740 Episode_Termination/fell_over: 0.0195 Episode_Termination/course_exit: 0.4935 Episode_Termination/illegal_nonwheel_contact: 0.0195 ============================================================================== ============================================================================== Learning iteration 26750/50000 Total steps: 109,568,000 Steps per second: 12,739 Iteration time: 0.322s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1392.4 Actor loss: -9.4797 Critic loss: +0.0081 World model loss: +0.6438 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2729 Episode_Reward/track_angular_velocity: +52.5701 Episode_Reward/upright: +26.7119 Episode_Reward/leg_posture: -2.6279 Episode_Reward/standstill: -6.0579 Episode_Reward/action_rate_l2: -0.6052 Episode_Reward/joint_torques_l2: -0.1520 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0245 Episode_Reward/unnecessary_wheel_lift: -0.0916 Episode_Reward/rolling_leg_motion: -0.6304 Episode_Reward/obstacle_progress: +3.6214 Episode_Reward/wheel_body_intrusion: -0.2863 Episode_Reward/pair_body_intrusion: -0.3602 Curriculum/terrain_level_mean: +3.6980 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.5094 Episode_Termination/fell_over: 0.0377 Episode_Termination/course_exit: 0.4340 Episode_Termination/illegal_nonwheel_contact: 0.0503 ============================================================================== ============================================================================== Learning iteration 26800/50000 Total steps: 109,772,800 Steps per second: 12,744 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1387.0 Actor loss: -9.5053 Critic loss: +0.0052 World model loss: +0.6429 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1848 Episode_Reward/track_angular_velocity: +52.4685 Episode_Reward/upright: +26.6415 Episode_Reward/leg_posture: -2.6523 Episode_Reward/standstill: -5.9316 Episode_Reward/action_rate_l2: -0.5822 Episode_Reward/joint_torques_l2: -0.1514 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0712 Episode_Reward/rolling_leg_motion: -0.6083 Episode_Reward/obstacle_progress: +3.6842 Episode_Reward/wheel_body_intrusion: -0.2853 Episode_Reward/pair_body_intrusion: -0.3416 Curriculum/terrain_level_mean: +3.7012 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4922 Episode_Termination/fell_over: 0.0078 Episode_Termination/course_exit: 0.4844 Episode_Termination/illegal_nonwheel_contact: 0.0234 ============================================================================== ============================================================================== Learning iteration 26850/50000 Total steps: 109,977,600 Steps per second: 12,748 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0871 Mean episode length: 1370.3 Actor loss: -9.5136 Critic loss: +0.0036 World model loss: +0.6855 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4609 Episode_Reward/track_angular_velocity: +51.7958 Episode_Reward/upright: +26.3116 Episode_Reward/leg_posture: -2.5972 Episode_Reward/standstill: -6.1275 Episode_Reward/action_rate_l2: -0.5770 Episode_Reward/joint_torques_l2: -0.1482 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.6042 Episode_Reward/obstacle_progress: +3.5863 Episode_Reward/wheel_body_intrusion: -0.2784 Episode_Reward/pair_body_intrusion: -0.3039 Curriculum/terrain_level_mean: +3.7043 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4830 Episode_Termination/fell_over: 0.0340 Episode_Termination/course_exit: 0.4694 Episode_Termination/illegal_nonwheel_contact: 0.0272 ============================================================================== ============================================================================== Learning iteration 26900/50000 Total steps: 110,182,400 Steps per second: 12,753 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0869 Mean episode length: 1338.1 Actor loss: -9.5061 Critic loss: +0.0076 World model loss: +0.6657 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3280 Episode_Reward/track_angular_velocity: +50.5304 Episode_Reward/upright: +25.6707 Episode_Reward/leg_posture: -2.5266 Episode_Reward/standstill: -6.2325 Episode_Reward/action_rate_l2: -0.5866 Episode_Reward/joint_torques_l2: -0.1457 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0751 Episode_Reward/rolling_leg_motion: -0.5913 Episode_Reward/obstacle_progress: +3.5002 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.2893 Curriculum/terrain_level_mean: +3.7068 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0423 Episode_Termination/course_exit: 0.4437 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== ============================================================================== Learning iteration 26950/50000 Total steps: 110,387,200 Steps per second: 12,757 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1416.5 Actor loss: -9.5146 Critic loss: +0.0037 World model loss: +0.6566 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1122 Episode_Reward/track_angular_velocity: +53.5230 Episode_Reward/upright: +27.1960 Episode_Reward/leg_posture: -2.6723 Episode_Reward/standstill: -5.8870 Episode_Reward/action_rate_l2: -0.6120 Episode_Reward/joint_torques_l2: -0.1525 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0055 Episode_Reward/unnecessary_wheel_lift: -0.0821 Episode_Reward/rolling_leg_motion: -0.6334 Episode_Reward/obstacle_progress: +3.7615 Episode_Reward/wheel_body_intrusion: -0.2563 Episode_Reward/pair_body_intrusion: -0.3481 Curriculum/terrain_level_mean: +3.7109 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.5373 Episode_Termination/fell_over: 0.0149 Episode_Termination/course_exit: 0.4478 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1612] ============================================================================== Learning iteration 27000/50000 Total steps: 110,592,000 Steps per second: 12,761 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0876 Mean episode length: 1401.2 Actor loss: -9.4941 Critic loss: +0.0053 World model loss: +0.6974 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6659 Episode_Reward/track_angular_velocity: +52.9634 Episode_Reward/upright: +26.9322 Episode_Reward/leg_posture: -2.6384 Episode_Reward/standstill: -5.8486 Episode_Reward/action_rate_l2: -0.6023 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0192 Episode_Reward/unnecessary_wheel_lift: -0.1021 Episode_Reward/rolling_leg_motion: -0.6232 Episode_Reward/obstacle_progress: +3.8054 Episode_Reward/wheel_body_intrusion: -0.2675 Episode_Reward/pair_body_intrusion: -0.3157 Curriculum/terrain_level_mean: +3.7139 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.5417 Episode_Termination/fell_over: 0.0417 Episode_Termination/course_exit: 0.4097 Episode_Termination/illegal_nonwheel_contact: 0.0417 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_027000.mp4 RECOVERY reset 1 invalid environments; first IDs=[2603] ============================================================================== Learning iteration 27050/50000 Total steps: 110,796,800 Steps per second: 12,757 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1451.7 Actor loss: -9.5092 Critic loss: +0.0054 World model loss: +0.6410 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.4013 Episode_Reward/track_angular_velocity: +54.8325 Episode_Reward/upright: +27.8639 Episode_Reward/leg_posture: -2.7824 Episode_Reward/standstill: -5.8016 Episode_Reward/action_rate_l2: -0.6125 Episode_Reward/joint_torques_l2: -0.1595 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0113 Episode_Reward/unnecessary_wheel_lift: -0.0914 Episode_Reward/rolling_leg_motion: -0.6518 Episode_Reward/obstacle_progress: +3.8226 Episode_Reward/wheel_body_intrusion: -0.2902 Episode_Reward/pair_body_intrusion: -0.3225 Curriculum/terrain_level_mean: +3.7166 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5674 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.4043 Episode_Termination/illegal_nonwheel_contact: 0.0142 ============================================================================== ============================================================================== Learning iteration 27100/50000 Total steps: 111,001,600 Steps per second: 12,761 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1438.6 Actor loss: -9.4866 Critic loss: +0.0048 World model loss: +0.6499 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0188 Episode_Reward/track_angular_velocity: +54.3522 Episode_Reward/upright: +27.6273 Episode_Reward/leg_posture: -2.7928 Episode_Reward/standstill: -5.9295 Episode_Reward/action_rate_l2: -0.6266 Episode_Reward/joint_torques_l2: -0.1568 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0140 Episode_Reward/unnecessary_wheel_lift: -0.0962 Episode_Reward/rolling_leg_motion: -0.6668 Episode_Reward/obstacle_progress: +3.8123 Episode_Reward/wheel_body_intrusion: -0.2890 Episode_Reward/pair_body_intrusion: -0.3184 Curriculum/terrain_level_mean: +3.7190 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5161 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4516 Episode_Termination/illegal_nonwheel_contact: 0.0258 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1398] ============================================================================== Learning iteration 27150/50000 Total steps: 111,206,400 Steps per second: 12,766 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0873 Mean episode length: 1350.3 Actor loss: -9.5029 Critic loss: +0.0029 World model loss: +0.6422 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7993 Episode_Reward/track_angular_velocity: +51.0015 Episode_Reward/upright: +25.9361 Episode_Reward/leg_posture: -2.5859 Episode_Reward/standstill: -5.8239 Episode_Reward/action_rate_l2: -0.5729 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: +0.0000 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.6185 Episode_Reward/obstacle_progress: +3.5350 Episode_Reward/wheel_body_intrusion: -0.2751 Episode_Reward/pair_body_intrusion: -0.3142 Curriculum/terrain_level_mean: +3.7234 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.3986 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.6014 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 27200/50000 Total steps: 111,411,200 Steps per second: 12,770 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0876 Mean episode length: 1334.4 Actor loss: -9.5037 Critic loss: +0.0029 World model loss: +0.6474 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3732 Episode_Reward/track_angular_velocity: +50.4444 Episode_Reward/upright: +25.6437 Episode_Reward/leg_posture: -2.5871 Episode_Reward/standstill: -5.4414 Episode_Reward/action_rate_l2: -0.5647 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5971 Episode_Reward/obstacle_progress: +3.5189 Episode_Reward/wheel_body_intrusion: -0.2734 Episode_Reward/pair_body_intrusion: -0.3009 Curriculum/terrain_level_mean: +3.7253 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4726 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4658 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2711] ============================================================================== Learning iteration 27250/50000 Total steps: 111,616,000 Steps per second: 12,775 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1447.3 Actor loss: -9.5080 Critic loss: +0.0034 World model loss: +0.6678 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.3842 Episode_Reward/track_angular_velocity: +54.6676 Episode_Reward/upright: +27.8272 Episode_Reward/leg_posture: -2.7642 Episode_Reward/standstill: -6.1378 Episode_Reward/action_rate_l2: -0.6231 Episode_Reward/joint_torques_l2: -0.1565 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0141 Episode_Reward/unnecessary_wheel_lift: -0.0859 Episode_Reward/rolling_leg_motion: -0.6639 Episode_Reward/obstacle_progress: +3.8448 Episode_Reward/wheel_body_intrusion: -0.2725 Episode_Reward/pair_body_intrusion: -0.3014 Curriculum/terrain_level_mean: +3.7295 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5759 Episode_Termination/fell_over: 0.0127 Episode_Termination/course_exit: 0.3924 Episode_Termination/illegal_nonwheel_contact: 0.0190 ============================================================================== ============================================================================== Learning iteration 27300/50000 Total steps: 111,820,800 Steps per second: 12,780 Iteration time: 0.321s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0873 Mean episode length: 1425.2 Actor loss: -9.4637 Critic loss: +0.0041 World model loss: +0.6563 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.4689 Episode_Reward/track_angular_velocity: +53.8409 Episode_Reward/upright: +27.3858 Episode_Reward/leg_posture: -2.7621 Episode_Reward/standstill: -6.1851 Episode_Reward/action_rate_l2: -0.6014 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0167 Episode_Reward/unnecessary_wheel_lift: -0.0862 Episode_Reward/rolling_leg_motion: -0.6431 Episode_Reward/obstacle_progress: +3.8205 Episode_Reward/wheel_body_intrusion: -0.2838 Episode_Reward/pair_body_intrusion: -0.3095 Curriculum/terrain_level_mean: +3.7310 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5427 Episode_Termination/fell_over: 0.0183 Episode_Termination/course_exit: 0.4207 Episode_Termination/illegal_nonwheel_contact: 0.0366 ============================================================================== ============================================================================== Learning iteration 27350/50000 Total steps: 112,025,600 Steps per second: 12,784 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1374.1 Actor loss: -9.5266 Critic loss: +0.0031 World model loss: +0.6722 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6507 Episode_Reward/track_angular_velocity: +51.9427 Episode_Reward/upright: +26.4121 Episode_Reward/leg_posture: -2.6158 Episode_Reward/standstill: -5.9664 Episode_Reward/action_rate_l2: -0.5868 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0242 Episode_Reward/unnecessary_wheel_lift: -0.0912 Episode_Reward/rolling_leg_motion: -0.6126 Episode_Reward/obstacle_progress: +3.6841 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.2882 Curriculum/terrain_level_mean: +3.7336 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4762 Episode_Termination/fell_over: 0.0476 Episode_Termination/course_exit: 0.4626 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 27400/50000 Total steps: 112,230,400 Steps per second: 12,788 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1378.3 Actor loss: -9.5409 Critic loss: +0.0033 World model loss: +0.6467 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7774 Episode_Reward/track_angular_velocity: +52.0286 Episode_Reward/upright: +26.4699 Episode_Reward/leg_posture: -2.6174 Episode_Reward/standstill: -5.9051 Episode_Reward/action_rate_l2: -0.5867 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0777 Episode_Reward/rolling_leg_motion: -0.5991 Episode_Reward/obstacle_progress: +3.6973 Episode_Reward/wheel_body_intrusion: -0.2650 Episode_Reward/pair_body_intrusion: -0.3073 Curriculum/terrain_level_mean: +3.7371 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.5188 Episode_Termination/fell_over: 0.0226 Episode_Termination/course_exit: 0.4511 Episode_Termination/illegal_nonwheel_contact: 0.0150 ============================================================================== ============================================================================== Learning iteration 27450/50000 Total steps: 112,435,200 Steps per second: 12,793 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1436.2 Actor loss: -9.4930 Critic loss: +0.0051 World model loss: +0.6464 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.8643 Episode_Reward/track_angular_velocity: +54.3101 Episode_Reward/upright: +27.5525 Episode_Reward/leg_posture: -2.7614 Episode_Reward/standstill: -6.2450 Episode_Reward/action_rate_l2: -0.6179 Episode_Reward/joint_torques_l2: -0.1566 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0140 Episode_Reward/unnecessary_wheel_lift: -0.0982 Episode_Reward/rolling_leg_motion: -0.6445 Episode_Reward/obstacle_progress: +3.8448 Episode_Reward/wheel_body_intrusion: -0.2795 Episode_Reward/pair_body_intrusion: -0.3181 Curriculum/terrain_level_mean: +3.7383 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5746 Episode_Termination/fell_over: 0.0149 Episode_Termination/course_exit: 0.3881 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 27500/50000 Total steps: 112,640,000 Steps per second: 12,797 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1368.2 Actor loss: -9.5152 Critic loss: +0.0032 World model loss: +0.6693 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4718 Episode_Reward/track_angular_velocity: +51.7552 Episode_Reward/upright: +26.2656 Episode_Reward/leg_posture: -2.6141 Episode_Reward/standstill: -5.6175 Episode_Reward/action_rate_l2: -0.5849 Episode_Reward/joint_torques_l2: -0.1503 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0053 Episode_Reward/unnecessary_wheel_lift: -0.0786 Episode_Reward/rolling_leg_motion: -0.6054 Episode_Reward/obstacle_progress: +3.6865 Episode_Reward/wheel_body_intrusion: -0.2630 Episode_Reward/pair_body_intrusion: -0.2764 Curriculum/terrain_level_mean: +3.7419 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0069 Episode_Termination/course_exit: 0.4828 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 27550/50000 Total steps: 112,844,800 Steps per second: 12,801 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0877 Mean episode length: 1358.6 Actor loss: -9.5101 Critic loss: +0.0034 World model loss: +0.6233 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1982 Episode_Reward/track_angular_velocity: +51.3582 Episode_Reward/upright: +26.1018 Episode_Reward/leg_posture: -2.5683 Episode_Reward/standstill: -5.6339 Episode_Reward/action_rate_l2: -0.5839 Episode_Reward/joint_torques_l2: -0.1468 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0821 Episode_Reward/rolling_leg_motion: -0.6117 Episode_Reward/obstacle_progress: +3.6666 Episode_Reward/wheel_body_intrusion: -0.2572 Episode_Reward/pair_body_intrusion: -0.2691 Curriculum/terrain_level_mean: +3.7458 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4717 Episode_Termination/fell_over: 0.0063 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0252 ============================================================================== ============================================================================== Learning iteration 27600/50000 Total steps: 113,049,600 Steps per second: 12,805 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0876 Mean episode length: 1349.2 Actor loss: -9.5203 Critic loss: +0.0045 World model loss: +0.6841 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7457 Episode_Reward/track_angular_velocity: +50.9893 Episode_Reward/upright: +25.9038 Episode_Reward/leg_posture: -2.5951 Episode_Reward/standstill: -5.3924 Episode_Reward/action_rate_l2: -0.5893 Episode_Reward/joint_torques_l2: -0.1461 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0237 Episode_Reward/unnecessary_wheel_lift: -0.0826 Episode_Reward/rolling_leg_motion: -0.6038 Episode_Reward/obstacle_progress: +3.5950 Episode_Reward/wheel_body_intrusion: -0.2698 Episode_Reward/pair_body_intrusion: -0.2899 Curriculum/terrain_level_mean: +3.7488 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4324 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.5068 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1413] ============================================================================== Learning iteration 27650/50000 Total steps: 113,254,400 Steps per second: 12,809 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1322.5 Actor loss: -9.5107 Critic loss: +0.0046 World model loss: +0.6709 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7169 Episode_Reward/track_angular_velocity: +49.9353 Episode_Reward/upright: +25.4192 Episode_Reward/leg_posture: -2.5271 Episode_Reward/standstill: -5.3171 Episode_Reward/action_rate_l2: -0.5907 Episode_Reward/joint_torques_l2: -0.1449 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.0984 Episode_Reward/rolling_leg_motion: -0.6264 Episode_Reward/obstacle_progress: +3.5023 Episode_Reward/wheel_body_intrusion: -0.2575 Episode_Reward/pair_body_intrusion: -0.2814 Curriculum/terrain_level_mean: +3.7517 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4036 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.5542 Episode_Termination/illegal_nonwheel_contact: 0.0120 ============================================================================== ============================================================================== Learning iteration 27700/50000 Total steps: 113,459,200 Steps per second: 12,814 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1365.2 Actor loss: -9.5054 Critic loss: +0.0029 World model loss: +0.6419 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3230 Episode_Reward/track_angular_velocity: +51.5562 Episode_Reward/upright: +26.2334 Episode_Reward/leg_posture: -2.6076 Episode_Reward/standstill: -5.7971 Episode_Reward/action_rate_l2: -0.5982 Episode_Reward/joint_torques_l2: -0.1498 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.0902 Episode_Reward/rolling_leg_motion: -0.6302 Episode_Reward/obstacle_progress: +3.6510 Episode_Reward/wheel_body_intrusion: -0.2741 Episode_Reward/pair_body_intrusion: -0.2946 Curriculum/terrain_level_mean: +3.7539 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4500 Episode_Termination/fell_over: 0.0312 Episode_Termination/course_exit: 0.4875 Episode_Termination/illegal_nonwheel_contact: 0.0500 ============================================================================== ============================================================================== Learning iteration 27750/50000 Total steps: 113,664,000 Steps per second: 12,818 Iteration time: 0.320s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1398.6 Actor loss: -9.5085 Critic loss: +0.0048 World model loss: +0.6407 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5863 Episode_Reward/track_angular_velocity: +52.8444 Episode_Reward/upright: +26.8954 Episode_Reward/leg_posture: -2.6545 Episode_Reward/standstill: -6.0506 Episode_Reward/action_rate_l2: -0.6099 Episode_Reward/joint_torques_l2: -0.1511 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0845 Episode_Reward/rolling_leg_motion: -0.6245 Episode_Reward/obstacle_progress: +3.7899 Episode_Reward/wheel_body_intrusion: -0.2820 Episode_Reward/pair_body_intrusion: -0.2823 Curriculum/terrain_level_mean: +3.7561 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5031 Episode_Termination/fell_over: 0.0252 Episode_Termination/course_exit: 0.4654 Episode_Termination/illegal_nonwheel_contact: 0.0252 ============================================================================== ============================================================================== Learning iteration 27800/50000 Total steps: 113,868,800 Steps per second: 12,822 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1378.4 Actor loss: -9.4902 Critic loss: +0.0088 World model loss: +0.6450 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8390 Episode_Reward/track_angular_velocity: +52.1336 Episode_Reward/upright: +26.4905 Episode_Reward/leg_posture: -2.6001 Episode_Reward/standstill: -6.1515 Episode_Reward/action_rate_l2: -0.5940 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0861 Episode_Reward/rolling_leg_motion: -0.6123 Episode_Reward/obstacle_progress: +3.6892 Episode_Reward/wheel_body_intrusion: -0.2636 Episode_Reward/pair_body_intrusion: -0.3023 Curriculum/terrain_level_mean: +3.7588 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5266 Episode_Termination/fell_over: 0.0237 Episode_Termination/course_exit: 0.4320 Episode_Termination/illegal_nonwheel_contact: 0.0237 ============================================================================== ============================================================================== Learning iteration 27850/50000 Total steps: 114,073,600 Steps per second: 12,826 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1435.7 Actor loss: -9.4975 Critic loss: +0.0057 World model loss: +0.6653 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.8179 Episode_Reward/track_angular_velocity: +54.2074 Episode_Reward/upright: +27.5656 Episode_Reward/leg_posture: -2.6758 Episode_Reward/standstill: -5.6683 Episode_Reward/action_rate_l2: -0.6114 Episode_Reward/joint_torques_l2: -0.1558 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0196 Episode_Reward/unnecessary_wheel_lift: -0.0883 Episode_Reward/rolling_leg_motion: -0.6469 Episode_Reward/obstacle_progress: +3.8380 Episode_Reward/wheel_body_intrusion: -0.2694 Episode_Reward/pair_body_intrusion: -0.2742 Curriculum/terrain_level_mean: +3.7607 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.5484 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4065 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2711] ============================================================================== Learning iteration 27900/50000 Total steps: 114,278,400 Steps per second: 12,830 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0877 Mean episode length: 1424.3 Actor loss: -9.5112 Critic loss: +0.0048 World model loss: +0.6470 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3764 Episode_Reward/track_angular_velocity: +53.8268 Episode_Reward/upright: +27.3384 Episode_Reward/leg_posture: -2.6586 Episode_Reward/standstill: -5.8320 Episode_Reward/action_rate_l2: -0.6082 Episode_Reward/joint_torques_l2: -0.1563 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0167 Episode_Reward/unnecessary_wheel_lift: -0.0960 Episode_Reward/rolling_leg_motion: -0.6322 Episode_Reward/obstacle_progress: +3.8724 Episode_Reward/wheel_body_intrusion: -0.2635 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +3.7620 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.5362 Episode_Termination/fell_over: 0.0217 Episode_Termination/course_exit: 0.4203 Episode_Termination/illegal_nonwheel_contact: 0.0435 ============================================================================== ============================================================================== Learning iteration 27950/50000 Total steps: 114,483,200 Steps per second: 12,835 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1331.1 Actor loss: -9.5093 Critic loss: +0.0042 World model loss: +0.6706 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0838 Episode_Reward/track_angular_velocity: +50.2903 Episode_Reward/upright: +25.5540 Episode_Reward/leg_posture: -2.5026 Episode_Reward/standstill: -5.7571 Episode_Reward/action_rate_l2: -0.5674 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0312 Episode_Reward/unnecessary_wheel_lift: -0.0848 Episode_Reward/rolling_leg_motion: -0.5876 Episode_Reward/obstacle_progress: +3.6314 Episode_Reward/wheel_body_intrusion: -0.2481 Episode_Reward/pair_body_intrusion: -0.2742 Curriculum/terrain_level_mean: +3.7649 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4667 Episode_Termination/fell_over: 0.0667 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0370 ============================================================================== ============================================================================== Learning iteration 28000/50000 Total steps: 114,688,000 Steps per second: 12,839 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1401.5 Actor loss: -9.4875 Critic loss: +0.0061 World model loss: +0.6450 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5057 Episode_Reward/track_angular_velocity: +52.9376 Episode_Reward/upright: +26.9288 Episode_Reward/leg_posture: -2.6473 Episode_Reward/standstill: -5.8255 Episode_Reward/action_rate_l2: -0.6052 Episode_Reward/joint_torques_l2: -0.1523 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0916 Episode_Reward/rolling_leg_motion: -0.6218 Episode_Reward/obstacle_progress: +3.8381 Episode_Reward/wheel_body_intrusion: -0.2795 Episode_Reward/pair_body_intrusion: -0.2954 Curriculum/terrain_level_mean: +3.7666 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5337 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.4356 Episode_Termination/illegal_nonwheel_contact: 0.0245 ============================================================================== ============================================================================== Learning iteration 28050/50000 Total steps: 114,892,800 Steps per second: 12,843 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1306.1 Actor loss: -9.5220 Critic loss: +0.0075 World model loss: +0.6258 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2083 Episode_Reward/track_angular_velocity: +49.3733 Episode_Reward/upright: +25.1091 Episode_Reward/leg_posture: -2.4595 Episode_Reward/standstill: -5.4119 Episode_Reward/action_rate_l2: -0.5548 Episode_Reward/joint_torques_l2: -0.1424 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0204 Episode_Reward/unnecessary_wheel_lift: -0.0842 Episode_Reward/rolling_leg_motion: -0.5819 Episode_Reward/obstacle_progress: +3.5373 Episode_Reward/wheel_body_intrusion: -0.2668 Episode_Reward/pair_body_intrusion: -0.3065 Curriculum/terrain_level_mean: +3.7688 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4228 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.5235 Episode_Termination/illegal_nonwheel_contact: 0.0336 ============================================================================== ============================================================================== Learning iteration 28100/50000 Total steps: 115,097,600 Steps per second: 12,847 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1411.5 Actor loss: -9.5225 Critic loss: +0.0039 World model loss: +0.6463 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9686 Episode_Reward/track_angular_velocity: +53.4313 Episode_Reward/upright: +27.1364 Episode_Reward/leg_posture: -2.6800 Episode_Reward/standstill: -5.9650 Episode_Reward/action_rate_l2: -0.5965 Episode_Reward/joint_torques_l2: -0.1523 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.0837 Episode_Reward/rolling_leg_motion: -0.6127 Episode_Reward/obstacle_progress: +3.7881 Episode_Reward/wheel_body_intrusion: -0.2734 Episode_Reward/pair_body_intrusion: -0.3000 Curriculum/terrain_level_mean: +3.7690 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.5347 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.4097 Episode_Termination/illegal_nonwheel_contact: 0.0347 ============================================================================== ============================================================================== Learning iteration 28150/50000 Total steps: 115,302,400 Steps per second: 12,851 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1313.2 Actor loss: -9.5277 Critic loss: +0.0050 World model loss: +0.6398 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5418 Episode_Reward/track_angular_velocity: +49.6471 Episode_Reward/upright: +25.2316 Episode_Reward/leg_posture: -2.4253 Episode_Reward/standstill: -5.5633 Episode_Reward/action_rate_l2: -0.5492 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0282 Episode_Reward/unnecessary_wheel_lift: -0.0798 Episode_Reward/rolling_leg_motion: -0.5668 Episode_Reward/obstacle_progress: +3.5381 Episode_Reward/wheel_body_intrusion: -0.2532 Episode_Reward/pair_body_intrusion: -0.2999 Curriculum/terrain_level_mean: +3.7710 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4568 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4753 Episode_Termination/illegal_nonwheel_contact: 0.0494 ============================================================================== ============================================================================== Learning iteration 28200/50000 Total steps: 115,507,200 Steps per second: 12,855 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1427.1 Actor loss: -9.4918 Critic loss: +0.0089 World model loss: +0.6503 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.4428 Episode_Reward/track_angular_velocity: +53.9271 Episode_Reward/upright: +27.3925 Episode_Reward/leg_posture: -2.7177 Episode_Reward/standstill: -6.1187 Episode_Reward/action_rate_l2: -0.5981 Episode_Reward/joint_torques_l2: -0.1570 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0167 Episode_Reward/unnecessary_wheel_lift: -0.0928 Episode_Reward/rolling_leg_motion: -0.6217 Episode_Reward/obstacle_progress: +3.7705 Episode_Reward/wheel_body_intrusion: -0.2971 Episode_Reward/pair_body_intrusion: -0.3667 Curriculum/terrain_level_mean: +3.7737 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4839 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4774 Episode_Termination/illegal_nonwheel_contact: 0.0258 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1868] ============================================================================== Learning iteration 28250/50000 Total steps: 115,712,000 Steps per second: 12,860 Iteration time: 0.319s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1370.7 Actor loss: -9.5288 Critic loss: +0.0037 World model loss: +0.6336 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5767 Episode_Reward/track_angular_velocity: +51.8483 Episode_Reward/upright: +26.3023 Episode_Reward/leg_posture: -2.5611 Episode_Reward/standstill: -5.8016 Episode_Reward/action_rate_l2: -0.5733 Episode_Reward/joint_torques_l2: -0.1501 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0791 Episode_Reward/rolling_leg_motion: -0.5853 Episode_Reward/obstacle_progress: +3.6607 Episode_Reward/wheel_body_intrusion: -0.2575 Episode_Reward/pair_body_intrusion: -0.2914 Curriculum/terrain_level_mean: +3.7764 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.4690 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 28300/50000 Total steps: 115,916,800 Steps per second: 12,864 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1321.7 Actor loss: -9.5227 Critic loss: +0.0043 World model loss: +0.6556 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8024 Episode_Reward/track_angular_velocity: +49.9782 Episode_Reward/upright: +25.3450 Episode_Reward/leg_posture: -2.5082 Episode_Reward/standstill: -5.4809 Episode_Reward/action_rate_l2: -0.5504 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.0733 Episode_Reward/rolling_leg_motion: -0.5629 Episode_Reward/obstacle_progress: +3.5753 Episode_Reward/wheel_body_intrusion: -0.2632 Episode_Reward/pair_body_intrusion: -0.2947 Curriculum/terrain_level_mean: +3.7781 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4514 Episode_Termination/fell_over: 0.0347 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0139 ============================================================================== ============================================================================== Learning iteration 28350/50000 Total steps: 116,121,600 Steps per second: 12,867 Iteration time: 0.318s Collection time: 0.075s Learning time: 0.192s Mean reward: +0.0877 Mean episode length: 1323.2 Actor loss: -9.5127 Critic loss: +0.0052 World model loss: +0.6642 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9299 Episode_Reward/track_angular_velocity: +49.9545 Episode_Reward/upright: +25.3973 Episode_Reward/leg_posture: -2.4732 Episode_Reward/standstill: -5.3606 Episode_Reward/action_rate_l2: -0.5642 Episode_Reward/joint_torques_l2: -0.1449 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.0758 Episode_Reward/rolling_leg_motion: -0.5802 Episode_Reward/obstacle_progress: +3.5473 Episode_Reward/wheel_body_intrusion: -0.2563 Episode_Reward/pair_body_intrusion: -0.3040 Curriculum/terrain_level_mean: +3.7815 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4167 Episode_Termination/fell_over: 0.0347 Episode_Termination/course_exit: 0.5278 Episode_Termination/illegal_nonwheel_contact: 0.0417 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1557] ============================================================================== Learning iteration 28400/50000 Total steps: 116,326,400 Steps per second: 12,871 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1329.3 Actor loss: -9.5428 Critic loss: +0.0044 World model loss: +0.6478 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0146 Episode_Reward/track_angular_velocity: +50.1461 Episode_Reward/upright: +25.5392 Episode_Reward/leg_posture: -2.4726 Episode_Reward/standstill: -5.5274 Episode_Reward/action_rate_l2: -0.5660 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0234 Episode_Reward/unnecessary_wheel_lift: -0.0986 Episode_Reward/rolling_leg_motion: -0.5861 Episode_Reward/obstacle_progress: +3.5747 Episode_Reward/wheel_body_intrusion: -0.2628 Episode_Reward/pair_body_intrusion: -0.2919 Curriculum/terrain_level_mean: +3.7830 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4398 Episode_Termination/fell_over: 0.0181 Episode_Termination/course_exit: 0.5060 Episode_Termination/illegal_nonwheel_contact: 0.0422 ============================================================================== ============================================================================== Learning iteration 28450/50000 Total steps: 116,531,200 Steps per second: 12,876 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1409.4 Actor loss: -9.5298 Critic loss: +0.0035 World model loss: +0.6324 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8764 Episode_Reward/track_angular_velocity: +53.2225 Episode_Reward/upright: +27.0638 Episode_Reward/leg_posture: -2.5941 Episode_Reward/standstill: -5.8981 Episode_Reward/action_rate_l2: -0.5809 Episode_Reward/joint_torques_l2: -0.1537 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.0948 Episode_Reward/rolling_leg_motion: -0.6184 Episode_Reward/obstacle_progress: +3.8155 Episode_Reward/wheel_body_intrusion: -0.2594 Episode_Reward/pair_body_intrusion: -0.2667 Curriculum/terrain_level_mean: +3.7849 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5096 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.4522 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[728] ============================================================================== Learning iteration 28500/50000 Total steps: 116,736,000 Steps per second: 12,880 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1407.0 Actor loss: -9.5493 Critic loss: +0.0049 World model loss: +0.6271 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7997 Episode_Reward/track_angular_velocity: +53.1302 Episode_Reward/upright: +26.9896 Episode_Reward/leg_posture: -2.6273 Episode_Reward/standstill: -5.7792 Episode_Reward/action_rate_l2: -0.5929 Episode_Reward/joint_torques_l2: -0.1540 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0247 Episode_Reward/unnecessary_wheel_lift: -0.0948 Episode_Reward/rolling_leg_motion: -0.6168 Episode_Reward/obstacle_progress: +3.8883 Episode_Reward/wheel_body_intrusion: -0.2710 Episode_Reward/pair_body_intrusion: -0.2845 Curriculum/terrain_level_mean: +3.7847 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5442 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.3946 Episode_Termination/illegal_nonwheel_contact: 0.0544 ============================================================================== ============================================================================== Learning iteration 28550/50000 Total steps: 116,940,800 Steps per second: 12,884 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1373.0 Actor loss: -9.4991 Critic loss: +0.0062 World model loss: +0.6257 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5238 Episode_Reward/track_angular_velocity: +51.8630 Episode_Reward/upright: +26.3474 Episode_Reward/leg_posture: -2.5662 Episode_Reward/standstill: -6.0235 Episode_Reward/action_rate_l2: -0.5878 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0768 Episode_Reward/rolling_leg_motion: -0.5884 Episode_Reward/obstacle_progress: +3.7727 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.3137 Curriculum/terrain_level_mean: +3.7866 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4714 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.4786 Episode_Termination/illegal_nonwheel_contact: 0.0357 ============================================================================== ============================================================================== Learning iteration 28600/50000 Total steps: 117,145,600 Steps per second: 12,888 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1293.1 Actor loss: -9.5166 Critic loss: +0.0056 World model loss: +0.6550 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7508 Episode_Reward/track_angular_velocity: +48.8055 Episode_Reward/upright: +24.7904 Episode_Reward/leg_posture: -2.3764 Episode_Reward/standstill: -5.3084 Episode_Reward/action_rate_l2: -0.5562 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0328 Episode_Reward/unnecessary_wheel_lift: -0.0931 Episode_Reward/rolling_leg_motion: -0.5707 Episode_Reward/obstacle_progress: +3.5728 Episode_Reward/wheel_body_intrusion: -0.2372 Episode_Reward/pair_body_intrusion: -0.2302 Curriculum/terrain_level_mean: +3.7905 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.3987 Episode_Termination/fell_over: 0.0506 Episode_Termination/course_exit: 0.5190 Episode_Termination/illegal_nonwheel_contact: 0.0570 ============================================================================== ============================================================================== Learning iteration 28650/50000 Total steps: 117,350,400 Steps per second: 12,892 Iteration time: 0.318s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1314.2 Actor loss: -9.5218 Critic loss: +0.0038 World model loss: +0.6547 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4690 Episode_Reward/track_angular_velocity: +49.6570 Episode_Reward/upright: +25.2309 Episode_Reward/leg_posture: -2.4265 Episode_Reward/standstill: -5.4154 Episode_Reward/action_rate_l2: -0.5629 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.0895 Episode_Reward/rolling_leg_motion: -0.5795 Episode_Reward/obstacle_progress: +3.5502 Episode_Reward/wheel_body_intrusion: -0.2661 Episode_Reward/pair_body_intrusion: -0.2724 Curriculum/terrain_level_mean: +3.7927 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4406 Episode_Termination/fell_over: 0.0210 Episode_Termination/course_exit: 0.4895 Episode_Termination/illegal_nonwheel_contact: 0.0559 ============================================================================== ============================================================================== Learning iteration 28700/50000 Total steps: 117,555,200 Steps per second: 12,896 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1305.6 Actor loss: -9.4998 Critic loss: +0.0048 World model loss: +0.6793 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2366 Episode_Reward/track_angular_velocity: +49.3407 Episode_Reward/upright: +25.0639 Episode_Reward/leg_posture: -2.4516 Episode_Reward/standstill: -5.2002 Episode_Reward/action_rate_l2: -0.5392 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0153 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.5707 Episode_Reward/obstacle_progress: +3.5135 Episode_Reward/wheel_body_intrusion: -0.2789 Episode_Reward/pair_body_intrusion: -0.2902 Curriculum/terrain_level_mean: +3.7949 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4583 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== ============================================================================== Learning iteration 28750/50000 Total steps: 117,760,000 Steps per second: 12,900 Iteration time: 0.318s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1393.8 Actor loss: -9.5355 Critic loss: +0.0050 World model loss: +0.6196 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2633 Episode_Reward/track_angular_velocity: +52.6261 Episode_Reward/upright: +26.7394 Episode_Reward/leg_posture: -2.6137 Episode_Reward/standstill: -5.9649 Episode_Reward/action_rate_l2: -0.5889 Episode_Reward/joint_torques_l2: -0.1529 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0913 Episode_Reward/rolling_leg_motion: -0.6109 Episode_Reward/obstacle_progress: +3.8072 Episode_Reward/wheel_body_intrusion: -0.2714 Episode_Reward/pair_body_intrusion: -0.2704 Curriculum/terrain_level_mean: +3.7976 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.4718 Episode_Termination/illegal_nonwheel_contact: 0.0141 ============================================================================== ============================================================================== Learning iteration 28800/50000 Total steps: 117,964,800 Steps per second: 12,904 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1361.8 Actor loss: -9.5451 Critic loss: +0.0039 World model loss: +0.6524 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1812 Episode_Reward/track_angular_velocity: +51.4131 Episode_Reward/upright: +26.0966 Episode_Reward/leg_posture: -2.5471 Episode_Reward/standstill: -5.8609 Episode_Reward/action_rate_l2: -0.5749 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.0933 Episode_Reward/rolling_leg_motion: -0.6054 Episode_Reward/obstacle_progress: +3.6595 Episode_Reward/wheel_body_intrusion: -0.2497 Episode_Reward/pair_body_intrusion: -0.2439 Curriculum/terrain_level_mean: +3.8005 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.4815 Episode_Termination/fell_over: 0.0309 Episode_Termination/course_exit: 0.4877 Episode_Termination/illegal_nonwheel_contact: 0.0185 ============================================================================== ============================================================================== Learning iteration 28850/50000 Total steps: 118,169,600 Steps per second: 12,908 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1402.5 Actor loss: -9.5295 Critic loss: +0.0051 World model loss: +0.6265 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8181 Episode_Reward/track_angular_velocity: +53.0251 Episode_Reward/upright: +26.9184 Episode_Reward/leg_posture: -2.6046 Episode_Reward/standstill: -6.0705 Episode_Reward/action_rate_l2: -0.5939 Episode_Reward/joint_torques_l2: -0.1531 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: +0.0000 Episode_Reward/unnecessary_wheel_lift: -0.0798 Episode_Reward/rolling_leg_motion: -0.6134 Episode_Reward/obstacle_progress: +3.7869 Episode_Reward/wheel_body_intrusion: -0.2676 Episode_Reward/pair_body_intrusion: -0.2711 Curriculum/terrain_level_mean: +3.8042 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.5203 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4797 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== ============================================================================== Learning iteration 28900/50000 Total steps: 118,374,400 Steps per second: 12,912 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1302.3 Actor loss: -9.5377 Critic loss: +0.0039 World model loss: +0.6439 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1969 Episode_Reward/track_angular_velocity: +49.1842 Episode_Reward/upright: +24.9610 Episode_Reward/leg_posture: -2.4201 Episode_Reward/standstill: -5.5272 Episode_Reward/action_rate_l2: -0.5569 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0153 Episode_Reward/unnecessary_wheel_lift: -0.0820 Episode_Reward/rolling_leg_motion: -0.5692 Episode_Reward/obstacle_progress: +3.5039 Episode_Reward/wheel_body_intrusion: -0.2454 Episode_Reward/pair_body_intrusion: -0.2461 Curriculum/terrain_level_mean: +3.8069 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.5306 Episode_Termination/illegal_nonwheel_contact: 0.0204 ============================================================================== ============================================================================== Learning iteration 28950/50000 Total steps: 118,579,200 Steps per second: 12,916 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1305.3 Actor loss: -9.5117 Critic loss: +0.0038 World model loss: +0.6847 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3211 Episode_Reward/track_angular_velocity: +49.3018 Episode_Reward/upright: +24.9895 Episode_Reward/leg_posture: -2.4856 Episode_Reward/standstill: -5.9100 Episode_Reward/action_rate_l2: -0.5577 Episode_Reward/joint_torques_l2: -0.1426 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0306 Episode_Reward/unnecessary_wheel_lift: -0.0818 Episode_Reward/rolling_leg_motion: -0.5784 Episode_Reward/obstacle_progress: +3.4787 Episode_Reward/wheel_body_intrusion: -0.2647 Episode_Reward/pair_body_intrusion: -0.2634 Curriculum/terrain_level_mean: +3.8071 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.3906 Episode_Termination/fell_over: 0.0625 Episode_Termination/course_exit: 0.5156 Episode_Termination/illegal_nonwheel_contact: 0.0625 ============================================================================== ============================================================================== Learning iteration 29000/50000 Total steps: 118,784,000 Steps per second: 12,919 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0874 Mean episode length: 1336.2 Actor loss: -9.5643 Critic loss: +0.0041 World model loss: +0.6651 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4199 Episode_Reward/track_angular_velocity: +50.4793 Episode_Reward/upright: +25.6574 Episode_Reward/leg_posture: -2.5537 Episode_Reward/standstill: -5.7427 Episode_Reward/action_rate_l2: -0.5636 Episode_Reward/joint_torques_l2: -0.1449 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0862 Episode_Reward/rolling_leg_motion: -0.5922 Episode_Reward/obstacle_progress: +3.5336 Episode_Reward/wheel_body_intrusion: -0.2679 Episode_Reward/pair_body_intrusion: -0.2983 Curriculum/terrain_level_mean: +3.8108 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4889 Episode_Termination/fell_over: 0.0222 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== ============================================================================== Learning iteration 29050/50000 Total steps: 118,988,800 Steps per second: 12,923 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1374.3 Actor loss: -9.5558 Critic loss: +0.0042 World model loss: +0.6314 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8039 Episode_Reward/track_angular_velocity: +51.9361 Episode_Reward/upright: +26.3961 Episode_Reward/leg_posture: -2.6142 Episode_Reward/standstill: -5.7201 Episode_Reward/action_rate_l2: -0.5847 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0081 Episode_Reward/unnecessary_wheel_lift: -0.0902 Episode_Reward/rolling_leg_motion: -0.6193 Episode_Reward/obstacle_progress: +3.6070 Episode_Reward/wheel_body_intrusion: -0.2713 Episode_Reward/pair_body_intrusion: -0.2779 Curriculum/terrain_level_mean: +3.8147 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4753 Episode_Termination/fell_over: 0.0062 Episode_Termination/course_exit: 0.5062 Episode_Termination/illegal_nonwheel_contact: 0.0123 ============================================================================== ============================================================================== Learning iteration 29100/50000 Total steps: 119,193,600 Steps per second: 12,927 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1336.0 Actor loss: -9.5228 Critic loss: +0.0056 World model loss: +0.6933 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5723 Episode_Reward/track_angular_velocity: +50.4593 Episode_Reward/upright: +25.6600 Episode_Reward/leg_posture: -2.4896 Episode_Reward/standstill: -5.4504 Episode_Reward/action_rate_l2: -0.5746 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0758 Episode_Reward/rolling_leg_motion: -0.5859 Episode_Reward/obstacle_progress: +3.6058 Episode_Reward/wheel_body_intrusion: -0.2512 Episode_Reward/pair_body_intrusion: -0.2740 Curriculum/terrain_level_mean: +3.8176 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4318 Episode_Termination/fell_over: 0.0114 Episode_Termination/course_exit: 0.5227 Episode_Termination/illegal_nonwheel_contact: 0.0341 ============================================================================== ============================================================================== Learning iteration 29150/50000 Total steps: 119,398,400 Steps per second: 12,931 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1319.6 Actor loss: -9.5184 Critic loss: +0.0054 World model loss: +0.6153 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8994 Episode_Reward/track_angular_velocity: +49.8521 Episode_Reward/upright: +25.3363 Episode_Reward/leg_posture: -2.5182 Episode_Reward/standstill: -5.7186 Episode_Reward/action_rate_l2: -0.5647 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0180 Episode_Reward/unnecessary_wheel_lift: -0.0787 Episode_Reward/rolling_leg_motion: -0.5783 Episode_Reward/obstacle_progress: +3.6197 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.2689 Curriculum/terrain_level_mean: +3.8213 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4331 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.5223 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1707] ============================================================================== Learning iteration 29200/50000 Total steps: 119,603,200 Steps per second: 12,935 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1387.5 Actor loss: -9.5292 Critic loss: +0.0031 World model loss: +0.6835 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2050 Episode_Reward/track_angular_velocity: +52.4351 Episode_Reward/upright: +26.6708 Episode_Reward/leg_posture: -2.6188 Episode_Reward/standstill: -5.9370 Episode_Reward/action_rate_l2: -0.6036 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.0751 Episode_Reward/rolling_leg_motion: -0.5949 Episode_Reward/obstacle_progress: +3.8066 Episode_Reward/wheel_body_intrusion: -0.2603 Episode_Reward/pair_body_intrusion: -0.3144 Curriculum/terrain_level_mean: +3.8220 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0462 Episode_Termination/course_exit: 0.4385 Episode_Termination/illegal_nonwheel_contact: 0.0231 ============================================================================== ============================================================================== Learning iteration 29250/50000 Total steps: 119,808,000 Steps per second: 12,939 Iteration time: 0.317s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1409.8 Actor loss: -9.5086 Critic loss: +0.0061 World model loss: +0.6019 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0119 Episode_Reward/track_angular_velocity: +53.3292 Episode_Reward/upright: +27.0354 Episode_Reward/leg_posture: -2.6476 Episode_Reward/standstill: -6.3504 Episode_Reward/action_rate_l2: -0.6112 Episode_Reward/joint_torques_l2: -0.1530 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0330 Episode_Reward/unnecessary_wheel_lift: -0.0854 Episode_Reward/rolling_leg_motion: -0.6225 Episode_Reward/obstacle_progress: +3.9063 Episode_Reward/wheel_body_intrusion: -0.2854 Episode_Reward/pair_body_intrusion: -0.3195 Curriculum/terrain_level_mean: +3.8225 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4967 Episode_Termination/fell_over: 0.0265 Episode_Termination/course_exit: 0.4238 Episode_Termination/illegal_nonwheel_contact: 0.0728 ============================================================================== ============================================================================== Learning iteration 29300/50000 Total steps: 120,012,800 Steps per second: 12,943 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1427.0 Actor loss: -9.5160 Critic loss: +0.0035 World model loss: +0.6464 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.6152 Episode_Reward/track_angular_velocity: +53.9835 Episode_Reward/upright: +27.3584 Episode_Reward/leg_posture: -2.6647 Episode_Reward/standstill: -6.3050 Episode_Reward/action_rate_l2: -0.6149 Episode_Reward/joint_torques_l2: -0.1552 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0139 Episode_Reward/unnecessary_wheel_lift: -0.0875 Episode_Reward/rolling_leg_motion: -0.6363 Episode_Reward/obstacle_progress: +3.9615 Episode_Reward/wheel_body_intrusion: -0.2797 Episode_Reward/pair_body_intrusion: -0.3148 Curriculum/terrain_level_mean: +3.8240 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5390 Episode_Termination/fell_over: 0.0071 Episode_Termination/course_exit: 0.4255 Episode_Termination/illegal_nonwheel_contact: 0.0284 ============================================================================== ============================================================================== Learning iteration 29350/50000 Total steps: 120,217,600 Steps per second: 12,946 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1317.0 Actor loss: -9.5057 Critic loss: +0.0071 World model loss: +0.6558 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5907 Episode_Reward/track_angular_velocity: +49.7579 Episode_Reward/upright: +25.2402 Episode_Reward/leg_posture: -2.5172 Episode_Reward/standstill: -5.6966 Episode_Reward/action_rate_l2: -0.5665 Episode_Reward/joint_torques_l2: -0.1436 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0257 Episode_Reward/unnecessary_wheel_lift: -0.0832 Episode_Reward/rolling_leg_motion: -0.5815 Episode_Reward/obstacle_progress: +3.6070 Episode_Reward/wheel_body_intrusion: -0.2732 Episode_Reward/pair_body_intrusion: -0.3010 Curriculum/terrain_level_mean: +3.8247 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.3831 Episode_Termination/fell_over: 0.0519 Episode_Termination/course_exit: 0.5519 Episode_Termination/illegal_nonwheel_contact: 0.0519 ============================================================================== ============================================================================== Learning iteration 29400/50000 Total steps: 120,422,400 Steps per second: 12,950 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1369.8 Actor loss: -9.5154 Critic loss: +0.0039 World model loss: +0.6609 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6480 Episode_Reward/track_angular_velocity: +51.6993 Episode_Reward/upright: +26.2908 Episode_Reward/leg_posture: -2.6173 Episode_Reward/standstill: -5.9881 Episode_Reward/action_rate_l2: -0.5646 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0844 Episode_Reward/rolling_leg_motion: -0.6282 Episode_Reward/obstacle_progress: +3.6521 Episode_Reward/wheel_body_intrusion: -0.2844 Episode_Reward/pair_body_intrusion: -0.2852 Curriculum/terrain_level_mean: +3.8286 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4690 Episode_Termination/fell_over: 0.0069 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0276 ============================================================================== ============================================================================== Learning iteration 29450/50000 Total steps: 120,627,200 Steps per second: 12,954 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1404.1 Actor loss: -9.5241 Critic loss: +0.0095 World model loss: +0.6148 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8753 Episode_Reward/track_angular_velocity: +53.1238 Episode_Reward/upright: +26.9589 Episode_Reward/leg_posture: -2.7216 Episode_Reward/standstill: -5.9516 Episode_Reward/action_rate_l2: -0.5763 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.0828 Episode_Reward/rolling_leg_motion: -0.6142 Episode_Reward/obstacle_progress: +3.6450 Episode_Reward/wheel_body_intrusion: -0.2916 Episode_Reward/pair_body_intrusion: -0.2543 Curriculum/terrain_level_mean: +3.8311 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5135 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.4324 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 29500/50000 Total steps: 120,832,000 Steps per second: 12,958 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0874 Mean episode length: 1380.9 Actor loss: -9.5106 Critic loss: +0.0111 World model loss: +0.6030 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0215 Episode_Reward/track_angular_velocity: +52.2497 Episode_Reward/upright: +26.4990 Episode_Reward/leg_posture: -2.6532 Episode_Reward/standstill: -6.0579 Episode_Reward/action_rate_l2: -0.5688 Episode_Reward/joint_torques_l2: -0.1486 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0805 Episode_Reward/rolling_leg_motion: -0.6040 Episode_Reward/obstacle_progress: +3.6115 Episode_Reward/wheel_body_intrusion: -0.2809 Episode_Reward/pair_body_intrusion: -0.2836 Curriculum/terrain_level_mean: +3.8352 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.5163 Episode_Termination/fell_over: 0.0327 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0131 ============================================================================== ============================================================================== Learning iteration 29550/50000 Total steps: 121,036,800 Steps per second: 12,962 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0874 Mean episode length: 1344.1 Actor loss: -9.5523 Critic loss: +0.0058 World model loss: +0.6345 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7401 Episode_Reward/track_angular_velocity: +50.8233 Episode_Reward/upright: +25.7901 Episode_Reward/leg_posture: -2.5705 Episode_Reward/standstill: -5.9693 Episode_Reward/action_rate_l2: -0.5623 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.0837 Episode_Reward/rolling_leg_motion: -0.5840 Episode_Reward/obstacle_progress: +3.5884 Episode_Reward/wheel_body_intrusion: -0.2801 Episode_Reward/pair_body_intrusion: -0.2811 Curriculum/terrain_level_mean: +3.8379 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4813 Episode_Termination/fell_over: 0.0312 Episode_Termination/course_exit: 0.4688 Episode_Termination/illegal_nonwheel_contact: 0.0312 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1014] ============================================================================== Learning iteration 29600/50000 Total steps: 121,241,600 Steps per second: 12,966 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1372.7 Actor loss: -9.5388 Critic loss: +0.0062 World model loss: +0.6018 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7063 Episode_Reward/track_angular_velocity: +51.8622 Episode_Reward/upright: +26.3784 Episode_Reward/leg_posture: -2.5650 Episode_Reward/standstill: -6.1328 Episode_Reward/action_rate_l2: -0.5791 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0793 Episode_Reward/rolling_leg_motion: -0.5993 Episode_Reward/obstacle_progress: +3.7507 Episode_Reward/wheel_body_intrusion: -0.2564 Episode_Reward/pair_body_intrusion: -0.2648 Curriculum/terrain_level_mean: +3.8411 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4895 Episode_Termination/fell_over: 0.0210 Episode_Termination/course_exit: 0.4895 Episode_Termination/illegal_nonwheel_contact: 0.0140 ============================================================================== ============================================================================== Learning iteration 29650/50000 Total steps: 121,446,400 Steps per second: 12,969 Iteration time: 0.316s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1324.3 Actor loss: -9.5388 Critic loss: +0.0042 World model loss: +0.6581 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8609 Episode_Reward/track_angular_velocity: +50.0294 Episode_Reward/upright: +25.4096 Episode_Reward/leg_posture: -2.5364 Episode_Reward/standstill: -6.2526 Episode_Reward/action_rate_l2: -0.5430 Episode_Reward/joint_torques_l2: -0.1452 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.0784 Episode_Reward/rolling_leg_motion: -0.5609 Episode_Reward/obstacle_progress: +3.6651 Episode_Reward/wheel_body_intrusion: -0.2684 Episode_Reward/pair_body_intrusion: -0.2894 Curriculum/terrain_level_mean: +3.8418 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4388 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4964 Episode_Termination/illegal_nonwheel_contact: 0.0576 ============================================================================== ============================================================================== Learning iteration 29700/50000 Total steps: 121,651,200 Steps per second: 12,973 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1294.1 Actor loss: -9.5155 Critic loss: +0.0039 World model loss: +0.6112 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9593 Episode_Reward/track_angular_velocity: +48.8788 Episode_Reward/upright: +24.8333 Episode_Reward/leg_posture: -2.4492 Episode_Reward/standstill: -5.8737 Episode_Reward/action_rate_l2: -0.5308 Episode_Reward/joint_torques_l2: -0.1419 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0177 Episode_Reward/unnecessary_wheel_lift: -0.0817 Episode_Reward/rolling_leg_motion: -0.5649 Episode_Reward/obstacle_progress: +3.6135 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.2711 Curriculum/terrain_level_mean: +3.8447 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.3958 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5556 Episode_Termination/illegal_nonwheel_contact: 0.0486 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_029700.mp4 RECOVERY reset 1 invalid environments; first IDs=[469] ============================================================================== Learning iteration 29750/50000 Total steps: 121,856,000 Steps per second: 12,969 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1425.0 Actor loss: -9.5024 Critic loss: +0.0038 World model loss: +0.6298 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.6367 Episode_Reward/track_angular_velocity: +53.8239 Episode_Reward/upright: +27.3589 Episode_Reward/leg_posture: -2.7535 Episode_Reward/standstill: -6.4607 Episode_Reward/action_rate_l2: -0.5855 Episode_Reward/joint_torques_l2: -0.1547 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0306 Episode_Reward/unnecessary_wheel_lift: -0.0778 Episode_Reward/rolling_leg_motion: -0.6026 Episode_Reward/obstacle_progress: +3.9827 Episode_Reward/wheel_body_intrusion: -0.3118 Episode_Reward/pair_body_intrusion: -0.2760 Curriculum/terrain_level_mean: +3.8442 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5342 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.3904 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 29800/50000 Total steps: 122,060,800 Steps per second: 12,973 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1290.2 Actor loss: -9.5318 Critic loss: +0.0056 World model loss: +0.6848 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7684 Episode_Reward/track_angular_velocity: +48.6833 Episode_Reward/upright: +24.7913 Episode_Reward/leg_posture: -2.4976 Episode_Reward/standstill: -6.2512 Episode_Reward/action_rate_l2: -0.5451 Episode_Reward/joint_torques_l2: -0.1404 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0302 Episode_Reward/unnecessary_wheel_lift: -0.0838 Episode_Reward/rolling_leg_motion: -0.5680 Episode_Reward/obstacle_progress: +3.6251 Episode_Reward/wheel_body_intrusion: -0.2849 Episode_Reward/pair_body_intrusion: -0.2964 Curriculum/terrain_level_mean: +3.8447 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.3892 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.5389 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2646] ============================================================================== Learning iteration 29850/50000 Total steps: 122,265,600 Steps per second: 12,976 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1401.7 Actor loss: -9.4984 Critic loss: +0.0049 World model loss: +0.6072 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8463 Episode_Reward/track_angular_velocity: +52.9953 Episode_Reward/upright: +26.9513 Episode_Reward/leg_posture: -2.6722 Episode_Reward/standstill: -6.4836 Episode_Reward/action_rate_l2: -0.5735 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0164 Episode_Reward/unnecessary_wheel_lift: -0.0816 Episode_Reward/rolling_leg_motion: -0.6047 Episode_Reward/obstacle_progress: +3.8874 Episode_Reward/wheel_body_intrusion: -0.2922 Episode_Reward/pair_body_intrusion: -0.3227 Curriculum/terrain_level_mean: +3.8455 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4833 Episode_Termination/fell_over: 0.0333 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0417 ============================================================================== ============================================================================== Learning iteration 29900/50000 Total steps: 122,470,400 Steps per second: 12,980 Iteration time: 0.316s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1395.7 Actor loss: -9.5441 Critic loss: +0.0053 World model loss: +0.6506 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5439 Episode_Reward/track_angular_velocity: +52.7450 Episode_Reward/upright: +26.8203 Episode_Reward/leg_posture: -2.6213 Episode_Reward/standstill: -6.2761 Episode_Reward/action_rate_l2: -0.5805 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.0889 Episode_Reward/rolling_leg_motion: -0.6065 Episode_Reward/obstacle_progress: +3.8379 Episode_Reward/wheel_body_intrusion: -0.2743 Episode_Reward/pair_body_intrusion: -0.3206 Curriculum/terrain_level_mean: +3.8477 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.5094 Episode_Termination/fell_over: 0.0252 Episode_Termination/course_exit: 0.4403 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 29950/50000 Total steps: 122,675,200 Steps per second: 12,984 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1340.1 Actor loss: -9.5294 Critic loss: +0.0071 World model loss: +0.6536 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6853 Episode_Reward/track_angular_velocity: +50.6951 Episode_Reward/upright: +25.7516 Episode_Reward/leg_posture: -2.5778 Episode_Reward/standstill: -5.8038 Episode_Reward/action_rate_l2: -0.5602 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0262 Episode_Reward/unnecessary_wheel_lift: -0.0934 Episode_Reward/rolling_leg_motion: -0.5889 Episode_Reward/obstacle_progress: +3.6862 Episode_Reward/wheel_body_intrusion: -0.2762 Episode_Reward/pair_body_intrusion: -0.2812 Curriculum/terrain_level_mean: +3.8499 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4083 Episode_Termination/fell_over: 0.0296 Episode_Termination/course_exit: 0.5325 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2435] ============================================================================== Learning iteration 30000/50000 Total steps: 122,880,000 Steps per second: 12,988 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1368.8 Actor loss: -9.5215 Critic loss: +0.0062 World model loss: +0.6238 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6589 Episode_Reward/track_angular_velocity: +51.7620 Episode_Reward/upright: +26.3066 Episode_Reward/leg_posture: -2.6437 Episode_Reward/standstill: -6.3180 Episode_Reward/action_rate_l2: -0.5806 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.0990 Episode_Reward/rolling_leg_motion: -0.6238 Episode_Reward/obstacle_progress: +3.6953 Episode_Reward/wheel_body_intrusion: -0.2706 Episode_Reward/pair_body_intrusion: -0.2874 Curriculum/terrain_level_mean: +3.8516 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4832 Episode_Termination/fell_over: 0.0268 Episode_Termination/course_exit: 0.4564 Episode_Termination/illegal_nonwheel_contact: 0.0470 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[976] RECOVERY reset 1 invalid environments; first IDs=[2363] ============================================================================== Learning iteration 30050/50000 Total steps: 123,084,800 Steps per second: 12,991 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1344.2 Actor loss: -9.5343 Critic loss: +0.0047 World model loss: +0.6710 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8723 Episode_Reward/track_angular_velocity: +50.7631 Episode_Reward/upright: +25.8192 Episode_Reward/leg_posture: -2.5557 Episode_Reward/standstill: -6.3274 Episode_Reward/action_rate_l2: -0.5784 Episode_Reward/joint_torques_l2: -0.1435 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0341 Episode_Reward/unnecessary_wheel_lift: -0.0943 Episode_Reward/rolling_leg_motion: -0.5996 Episode_Reward/obstacle_progress: +3.6853 Episode_Reward/wheel_body_intrusion: -0.2535 Episode_Reward/pair_body_intrusion: -0.2821 Curriculum/terrain_level_mean: +3.8518 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4324 Episode_Termination/fell_over: 0.0541 Episode_Termination/course_exit: 0.4797 Episode_Termination/illegal_nonwheel_contact: 0.0676 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1376] ============================================================================== Learning iteration 30100/50000 Total steps: 123,289,600 Steps per second: 12,995 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1333.0 Actor loss: -9.5352 Critic loss: +0.0031 World model loss: +0.6579 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2814 Episode_Reward/track_angular_velocity: +50.3562 Episode_Reward/upright: +25.5642 Episode_Reward/leg_posture: -2.5336 Episode_Reward/standstill: -6.3757 Episode_Reward/action_rate_l2: -0.5651 Episode_Reward/joint_torques_l2: -0.1430 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0208 Episode_Reward/unnecessary_wheel_lift: -0.1014 Episode_Reward/rolling_leg_motion: -0.6081 Episode_Reward/obstacle_progress: +3.5385 Episode_Reward/wheel_body_intrusion: -0.2647 Episode_Reward/pair_body_intrusion: -0.3394 Curriculum/terrain_level_mean: +3.8525 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4371 Episode_Termination/fell_over: 0.0265 Episode_Termination/course_exit: 0.5099 Episode_Termination/illegal_nonwheel_contact: 0.0464 ============================================================================== ============================================================================== Learning iteration 30150/50000 Total steps: 123,494,400 Steps per second: 12,999 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1377.9 Actor loss: -9.5253 Critic loss: +0.0066 World model loss: +0.6262 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9879 Episode_Reward/track_angular_velocity: +52.1034 Episode_Reward/upright: +26.4508 Episode_Reward/leg_posture: -2.6145 Episode_Reward/standstill: -6.4612 Episode_Reward/action_rate_l2: -0.5927 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0269 Episode_Reward/unnecessary_wheel_lift: -0.1137 Episode_Reward/rolling_leg_motion: -0.6283 Episode_Reward/obstacle_progress: +3.6842 Episode_Reward/wheel_body_intrusion: -0.2654 Episode_Reward/pair_body_intrusion: -0.3106 Curriculum/terrain_level_mean: +3.8542 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4615 Episode_Termination/fell_over: 0.0256 Episode_Termination/course_exit: 0.4744 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1522] RECOVERY reset 1 invalid environments; first IDs=[1790] ============================================================================== Learning iteration 30200/50000 Total steps: 123,699,200 Steps per second: 13,002 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1319.9 Actor loss: -9.5193 Critic loss: +0.0042 World model loss: +0.6318 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8188 Episode_Reward/track_angular_velocity: +49.8708 Episode_Reward/upright: +25.3343 Episode_Reward/leg_posture: -2.4788 Episode_Reward/standstill: -5.9336 Episode_Reward/action_rate_l2: -0.5631 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0309 Episode_Reward/unnecessary_wheel_lift: -0.0960 Episode_Reward/rolling_leg_motion: -0.5791 Episode_Reward/obstacle_progress: +3.5524 Episode_Reward/wheel_body_intrusion: -0.2484 Episode_Reward/pair_body_intrusion: -0.2785 Curriculum/terrain_level_mean: +3.8562 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4431 Episode_Termination/fell_over: 0.0299 Episode_Termination/course_exit: 0.4850 Episode_Termination/illegal_nonwheel_contact: 0.0539 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2002] ============================================================================== Learning iteration 30250/50000 Total steps: 123,904,000 Steps per second: 13,006 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1338.8 Actor loss: -9.5066 Critic loss: +0.0063 World model loss: +0.6004 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5064 Episode_Reward/track_angular_velocity: +50.6009 Episode_Reward/upright: +25.6792 Episode_Reward/leg_posture: -2.5063 Episode_Reward/standstill: -6.2087 Episode_Reward/action_rate_l2: -0.5757 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0314 Episode_Reward/unnecessary_wheel_lift: -0.0947 Episode_Reward/rolling_leg_motion: -0.6062 Episode_Reward/obstacle_progress: +3.5876 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.3050 Curriculum/terrain_level_mean: +3.8574 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4733 Episode_Termination/fell_over: 0.0467 Episode_Termination/course_exit: 0.4467 Episode_Termination/illegal_nonwheel_contact: 0.0533 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2056] RECOVERY reset 1 invalid environments; first IDs=[1700] ============================================================================== Learning iteration 30300/50000 Total steps: 124,108,800 Steps per second: 13,010 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1318.4 Actor loss: -9.5403 Critic loss: +0.0077 World model loss: +0.6081 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8942 Episode_Reward/track_angular_velocity: +49.8431 Episode_Reward/upright: +25.2623 Episode_Reward/leg_posture: -2.4989 Episode_Reward/standstill: -6.0911 Episode_Reward/action_rate_l2: -0.5530 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0808 Episode_Reward/rolling_leg_motion: -0.5738 Episode_Reward/obstacle_progress: +3.5382 Episode_Reward/wheel_body_intrusion: -0.2413 Episode_Reward/pair_body_intrusion: -0.2861 Curriculum/terrain_level_mean: +3.8584 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4748 Episode_Termination/fell_over: 0.0288 Episode_Termination/course_exit: 0.4676 Episode_Termination/illegal_nonwheel_contact: 0.0432 ============================================================================== ============================================================================== Learning iteration 30350/50000 Total steps: 124,313,600 Steps per second: 13,013 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0872 Mean episode length: 1380.0 Actor loss: -9.5013 Critic loss: +0.0126 World model loss: +0.6255 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0082 Episode_Reward/track_angular_velocity: +52.1324 Episode_Reward/upright: +26.4611 Episode_Reward/leg_posture: -2.6172 Episode_Reward/standstill: -6.3086 Episode_Reward/action_rate_l2: -0.5831 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0189 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.5891 Episode_Reward/obstacle_progress: +3.7197 Episode_Reward/wheel_body_intrusion: -0.2663 Episode_Reward/pair_body_intrusion: -0.3043 Curriculum/terrain_level_mean: +3.8611 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4774 Episode_Termination/fell_over: 0.0323 Episode_Termination/course_exit: 0.4774 Episode_Termination/illegal_nonwheel_contact: 0.0258 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2165] ============================================================================== Learning iteration 30400/50000 Total steps: 124,518,400 Steps per second: 13,016 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1314.6 Actor loss: -9.5180 Critic loss: +0.0044 World model loss: +0.6156 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7117 Episode_Reward/track_angular_velocity: +49.6773 Episode_Reward/upright: +25.2072 Episode_Reward/leg_posture: -2.4483 Episode_Reward/standstill: -6.0905 Episode_Reward/action_rate_l2: -0.5636 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0180 Episode_Reward/unnecessary_wheel_lift: -0.0861 Episode_Reward/rolling_leg_motion: -0.5957 Episode_Reward/obstacle_progress: +3.5512 Episode_Reward/wheel_body_intrusion: -0.2320 Episode_Reward/pair_body_intrusion: -0.2362 Curriculum/terrain_level_mean: +3.8630 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4207 Episode_Termination/fell_over: 0.0305 Episode_Termination/course_exit: 0.5366 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[829] ============================================================================== Learning iteration 30450/50000 Total steps: 124,723,200 Steps per second: 13,020 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1370.3 Actor loss: -9.5353 Critic loss: +0.0050 World model loss: +0.6206 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7183 Episode_Reward/track_angular_velocity: +51.7446 Episode_Reward/upright: +26.2962 Episode_Reward/leg_posture: -2.5773 Episode_Reward/standstill: -6.2448 Episode_Reward/action_rate_l2: -0.5677 Episode_Reward/joint_torques_l2: -0.1479 Episode_Reward/dof_pos_limits: -0.0019 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0924 Episode_Reward/rolling_leg_motion: -0.6078 Episode_Reward/obstacle_progress: +3.6601 Episode_Reward/wheel_body_intrusion: -0.2587 Episode_Reward/pair_body_intrusion: -0.2696 Curriculum/terrain_level_mean: +3.8650 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0203 ============================================================================== ============================================================================== Learning iteration 30500/50000 Total steps: 124,928,000 Steps per second: 13,024 Iteration time: 0.315s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1327.1 Actor loss: -9.5178 Critic loss: +0.0062 World model loss: +0.6357 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1265 Episode_Reward/track_angular_velocity: +50.1728 Episode_Reward/upright: +25.4418 Episode_Reward/leg_posture: -2.5568 Episode_Reward/standstill: -6.1537 Episode_Reward/action_rate_l2: -0.5614 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0078 Episode_Reward/unnecessary_wheel_lift: -0.1021 Episode_Reward/rolling_leg_motion: -0.5947 Episode_Reward/obstacle_progress: +3.5477 Episode_Reward/wheel_body_intrusion: -0.2713 Episode_Reward/pair_body_intrusion: -0.3068 Curriculum/terrain_level_mean: +3.8713 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4242 Episode_Termination/fell_over: 0.0121 Episode_Termination/course_exit: 0.5576 Episode_Termination/illegal_nonwheel_contact: 0.0182 ============================================================================== ============================================================================== Learning iteration 30550/50000 Total steps: 125,132,800 Steps per second: 13,027 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1372.2 Actor loss: -9.5308 Critic loss: +0.0055 World model loss: +0.6149 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7968 Episode_Reward/track_angular_velocity: +51.8866 Episode_Reward/upright: +26.3481 Episode_Reward/leg_posture: -2.6133 Episode_Reward/standstill: -6.0777 Episode_Reward/action_rate_l2: -0.5710 Episode_Reward/joint_torques_l2: -0.1486 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0792 Episode_Reward/rolling_leg_motion: -0.5859 Episode_Reward/obstacle_progress: +3.7127 Episode_Reward/wheel_body_intrusion: -0.2677 Episode_Reward/pair_body_intrusion: -0.2969 Curriculum/terrain_level_mean: +3.8723 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4846 Episode_Termination/fell_over: 0.0154 Episode_Termination/course_exit: 0.4615 Episode_Termination/illegal_nonwheel_contact: 0.0462 ============================================================================== ============================================================================== Learning iteration 30600/50000 Total steps: 125,337,600 Steps per second: 13,031 Iteration time: 0.314s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1369.9 Actor loss: -9.5123 Critic loss: +0.0040 World model loss: +0.5797 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7620 Episode_Reward/track_angular_velocity: +51.8300 Episode_Reward/upright: +26.3137 Episode_Reward/leg_posture: -2.5884 Episode_Reward/standstill: -6.2957 Episode_Reward/action_rate_l2: -0.5737 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0921 Episode_Reward/rolling_leg_motion: -0.5905 Episode_Reward/obstacle_progress: +3.7903 Episode_Reward/wheel_body_intrusion: -0.2603 Episode_Reward/pair_body_intrusion: -0.2750 Curriculum/terrain_level_mean: +3.8733 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5278 Episode_Termination/fell_over: 0.0069 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2407] ============================================================================== Learning iteration 30650/50000 Total steps: 125,542,400 Steps per second: 13,035 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1433.6 Actor loss: -9.5103 Critic loss: +0.0039 World model loss: +0.6319 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.9368 Episode_Reward/track_angular_velocity: +54.2074 Episode_Reward/upright: +27.4894 Episode_Reward/leg_posture: -2.7515 Episode_Reward/standstill: -6.5560 Episode_Reward/action_rate_l2: -0.5974 Episode_Reward/joint_torques_l2: -0.1555 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0084 Episode_Reward/unnecessary_wheel_lift: -0.0810 Episode_Reward/rolling_leg_motion: -0.6074 Episode_Reward/obstacle_progress: +4.0336 Episode_Reward/wheel_body_intrusion: -0.2979 Episode_Reward/pair_body_intrusion: -0.3359 Curriculum/terrain_level_mean: +3.8777 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5496 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4275 Episode_Termination/illegal_nonwheel_contact: 0.0229 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1638] ============================================================================== Learning iteration 30700/50000 Total steps: 125,747,200 Steps per second: 13,038 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1331.0 Actor loss: -9.5267 Critic loss: +0.0044 World model loss: +0.6104 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1715 Episode_Reward/track_angular_velocity: +50.3373 Episode_Reward/upright: +25.5262 Episode_Reward/leg_posture: -2.4880 Episode_Reward/standstill: -6.2925 Episode_Reward/action_rate_l2: -0.5548 Episode_Reward/joint_torques_l2: -0.1431 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.0783 Episode_Reward/rolling_leg_motion: -0.5595 Episode_Reward/obstacle_progress: +3.7100 Episode_Reward/wheel_body_intrusion: -0.2750 Episode_Reward/pair_body_intrusion: -0.2811 Curriculum/terrain_level_mean: +3.8801 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4113 Episode_Termination/fell_over: 0.0284 Episode_Termination/course_exit: 0.5532 Episode_Termination/illegal_nonwheel_contact: 0.0142 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2682] RECOVERY reset 1 invalid environments; first IDs=[1247] ============================================================================== Learning iteration 30750/50000 Total steps: 125,952,000 Steps per second: 13,042 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1443.6 Actor loss: -9.5376 Critic loss: +0.0036 World model loss: +0.6131 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.2973 Episode_Reward/track_angular_velocity: +54.6298 Episode_Reward/upright: +27.7180 Episode_Reward/leg_posture: -2.6943 Episode_Reward/standstill: -6.4608 Episode_Reward/action_rate_l2: -0.5995 Episode_Reward/joint_torques_l2: -0.1543 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0113 Episode_Reward/unnecessary_wheel_lift: -0.0939 Episode_Reward/rolling_leg_motion: -0.6190 Episode_Reward/obstacle_progress: +3.9526 Episode_Reward/wheel_body_intrusion: -0.2852 Episode_Reward/pair_body_intrusion: -0.3352 Curriculum/terrain_level_mean: +3.8835 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.5663 Episode_Termination/fell_over: 0.0120 Episode_Termination/course_exit: 0.4096 Episode_Termination/illegal_nonwheel_contact: 0.0120 ============================================================================== ============================================================================== Learning iteration 30800/50000 Total steps: 126,156,800 Steps per second: 13,046 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0872 Mean episode length: 1370.3 Actor loss: -9.4968 Critic loss: +0.0058 World model loss: +0.6191 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7255 Episode_Reward/track_angular_velocity: +51.9181 Episode_Reward/upright: +26.3080 Episode_Reward/leg_posture: -2.5920 Episode_Reward/standstill: -6.5882 Episode_Reward/action_rate_l2: -0.5567 Episode_Reward/joint_torques_l2: -0.1471 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0268 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.5827 Episode_Reward/obstacle_progress: +3.6772 Episode_Reward/wheel_body_intrusion: -0.2781 Episode_Reward/pair_body_intrusion: -0.3027 Curriculum/terrain_level_mean: +3.8860 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1313 Episode_Termination/time_out: 0.4731 Episode_Termination/fell_over: 0.0180 Episode_Termination/course_exit: 0.4671 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 30850/50000 Total steps: 126,361,600 Steps per second: 13,049 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1357.5 Actor loss: -9.5337 Critic loss: +0.0037 World model loss: +0.6266 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1880 Episode_Reward/track_angular_velocity: +51.4037 Episode_Reward/upright: +26.0481 Episode_Reward/leg_posture: -2.5512 Episode_Reward/standstill: -6.5186 Episode_Reward/action_rate_l2: -0.5780 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0398 Episode_Reward/unnecessary_wheel_lift: -0.1044 Episode_Reward/rolling_leg_motion: -0.6044 Episode_Reward/obstacle_progress: +3.6432 Episode_Reward/wheel_body_intrusion: -0.2521 Episode_Reward/pair_body_intrusion: -0.2895 Curriculum/terrain_level_mean: +3.8875 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1316 Episode_Termination/time_out: 0.4737 Episode_Termination/fell_over: 0.0468 Episode_Termination/course_exit: 0.4386 Episode_Termination/illegal_nonwheel_contact: 0.0643 ============================================================================== ============================================================================== Learning iteration 30900/50000 Total steps: 126,566,400 Steps per second: 13,053 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1333.2 Actor loss: -9.5142 Critic loss: +0.0045 World model loss: +0.6297 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5450 Episode_Reward/track_angular_velocity: +50.5009 Episode_Reward/upright: +25.6153 Episode_Reward/leg_posture: -2.5275 Episode_Reward/standstill: -6.1522 Episode_Reward/action_rate_l2: -0.5566 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0365 Episode_Reward/unnecessary_wheel_lift: -0.0982 Episode_Reward/rolling_leg_motion: -0.5978 Episode_Reward/obstacle_progress: +3.5696 Episode_Reward/wheel_body_intrusion: -0.2632 Episode_Reward/pair_body_intrusion: -0.3159 Curriculum/terrain_level_mean: +3.8879 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.4695 Episode_Termination/fell_over: 0.0366 Episode_Termination/course_exit: 0.4451 Episode_Termination/illegal_nonwheel_contact: 0.0610 ============================================================================== ============================================================================== Learning iteration 30950/50000 Total steps: 126,771,200 Steps per second: 13,057 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1336.4 Actor loss: -9.5084 Critic loss: +0.0050 World model loss: +0.6052 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7326 Episode_Reward/track_angular_velocity: +50.6776 Episode_Reward/upright: +25.6554 Episode_Reward/leg_posture: -2.5442 Episode_Reward/standstill: -6.0514 Episode_Reward/action_rate_l2: -0.5565 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0820 Episode_Reward/rolling_leg_motion: -0.5853 Episode_Reward/obstacle_progress: +3.6552 Episode_Reward/wheel_body_intrusion: -0.2684 Episode_Reward/pair_body_intrusion: -0.2871 Curriculum/terrain_level_mean: +3.8916 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0898 Metrics/base_velocity/error_vel_yaw: 0.1305 Episode_Termination/time_out: 0.4297 Episode_Termination/fell_over: 0.0234 Episode_Termination/course_exit: 0.5234 Episode_Termination/illegal_nonwheel_contact: 0.0312 ============================================================================== ============================================================================== Learning iteration 31000/50000 Total steps: 126,976,000 Steps per second: 13,060 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1335.8 Actor loss: -9.5177 Critic loss: +0.0050 World model loss: +0.6175 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4524 Episode_Reward/track_angular_velocity: +50.6276 Episode_Reward/upright: +25.6071 Episode_Reward/leg_posture: -2.5277 Episode_Reward/standstill: -6.1689 Episode_Reward/action_rate_l2: -0.5635 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0824 Episode_Reward/rolling_leg_motion: -0.5710 Episode_Reward/obstacle_progress: +3.6894 Episode_Reward/wheel_body_intrusion: -0.2550 Episode_Reward/pair_body_intrusion: -0.2876 Curriculum/terrain_level_mean: +3.8926 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.4610 Episode_Termination/fell_over: 0.0355 Episode_Termination/course_exit: 0.4752 Episode_Termination/illegal_nonwheel_contact: 0.0284 ============================================================================== ============================================================================== Learning iteration 31050/50000 Total steps: 127,180,800 Steps per second: 13,063 Iteration time: 0.314s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1338.4 Actor loss: -9.5126 Critic loss: +0.0059 World model loss: +0.6085 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4506 Episode_Reward/track_angular_velocity: +50.6800 Episode_Reward/upright: +25.6058 Episode_Reward/leg_posture: -2.5314 Episode_Reward/standstill: -6.0520 Episode_Reward/action_rate_l2: -0.5689 Episode_Reward/joint_torques_l2: -0.1438 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0831 Episode_Reward/rolling_leg_motion: -0.5730 Episode_Reward/obstacle_progress: +3.6156 Episode_Reward/wheel_body_intrusion: -0.2396 Episode_Reward/pair_body_intrusion: -0.2872 Curriculum/terrain_level_mean: +3.8950 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4161 Episode_Termination/fell_over: 0.0134 Episode_Termination/course_exit: 0.5570 Episode_Termination/illegal_nonwheel_contact: 0.0134 ============================================================================== ============================================================================== Learning iteration 31100/50000 Total steps: 127,385,600 Steps per second: 13,067 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1370.4 Actor loss: -9.4958 Critic loss: +0.0037 World model loss: +0.6000 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5066 Episode_Reward/track_angular_velocity: +51.8630 Episode_Reward/upright: +26.2134 Episode_Reward/leg_posture: -2.5679 Episode_Reward/standstill: -6.1003 Episode_Reward/action_rate_l2: -0.5740 Episode_Reward/joint_torques_l2: -0.1472 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.0819 Episode_Reward/rolling_leg_motion: -0.5778 Episode_Reward/obstacle_progress: +3.6344 Episode_Reward/wheel_body_intrusion: -0.2370 Episode_Reward/pair_body_intrusion: -0.2878 Curriculum/terrain_level_mean: +3.8975 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0408 Episode_Termination/course_exit: 0.4422 Episode_Termination/illegal_nonwheel_contact: 0.0136 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[967] ============================================================================== Learning iteration 31150/50000 Total steps: 127,590,400 Steps per second: 13,070 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1332.7 Actor loss: -9.5245 Critic loss: +0.0041 World model loss: +0.6226 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2469 Episode_Reward/track_angular_velocity: +50.4982 Episode_Reward/upright: +25.5064 Episode_Reward/leg_posture: -2.5568 Episode_Reward/standstill: -6.1010 Episode_Reward/action_rate_l2: -0.5546 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0052 Episode_Reward/unnecessary_wheel_lift: -0.0951 Episode_Reward/rolling_leg_motion: -0.5829 Episode_Reward/obstacle_progress: +3.5599 Episode_Reward/wheel_body_intrusion: -0.2495 Episode_Reward/pair_body_intrusion: -0.2978 Curriculum/terrain_level_mean: +3.9019 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.4730 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.5135 Episode_Termination/illegal_nonwheel_contact: 0.0135 ============================================================================== ============================================================================== Learning iteration 31200/50000 Total steps: 127,795,200 Steps per second: 13,074 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1324.1 Actor loss: -9.5122 Critic loss: +0.0049 World model loss: +0.6074 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0611 Episode_Reward/track_angular_velocity: +50.0919 Episode_Reward/upright: +25.3866 Episode_Reward/leg_posture: -2.5294 Episode_Reward/standstill: -6.1392 Episode_Reward/action_rate_l2: -0.5549 Episode_Reward/joint_torques_l2: -0.1425 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.0938 Episode_Reward/rolling_leg_motion: -0.5876 Episode_Reward/obstacle_progress: +3.5113 Episode_Reward/wheel_body_intrusion: -0.2456 Episode_Reward/pair_body_intrusion: -0.2518 Curriculum/terrain_level_mean: +3.9036 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.4932 Episode_Termination/illegal_nonwheel_contact: 0.0270 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2331] ============================================================================== Learning iteration 31250/50000 Total steps: 128,000,000 Steps per second: 13,077 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1304.9 Actor loss: -9.5225 Critic loss: +0.0040 World model loss: +0.6129 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4743 Episode_Reward/track_angular_velocity: +49.4209 Episode_Reward/upright: +25.0237 Episode_Reward/leg_posture: -2.4818 Episode_Reward/standstill: -5.9386 Episode_Reward/action_rate_l2: -0.5397 Episode_Reward/joint_torques_l2: -0.1402 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0127 Episode_Reward/unnecessary_wheel_lift: -0.0959 Episode_Reward/rolling_leg_motion: -0.5727 Episode_Reward/obstacle_progress: +3.4989 Episode_Reward/wheel_body_intrusion: -0.2438 Episode_Reward/pair_body_intrusion: -0.3044 Curriculum/terrain_level_mean: +3.9062 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.4451 Episode_Termination/fell_over: 0.0183 Episode_Termination/course_exit: 0.5244 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[3279] ============================================================================== Learning iteration 31300/50000 Total steps: 128,204,800 Steps per second: 13,081 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1392.2 Actor loss: -9.4939 Critic loss: +0.0045 World model loss: +0.6026 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5104 Episode_Reward/track_angular_velocity: +52.7272 Episode_Reward/upright: +26.7166 Episode_Reward/leg_posture: -2.6786 Episode_Reward/standstill: -6.5641 Episode_Reward/action_rate_l2: -0.5742 Episode_Reward/joint_torques_l2: -0.1499 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.0883 Episode_Reward/rolling_leg_motion: -0.6106 Episode_Reward/obstacle_progress: +3.6671 Episode_Reward/wheel_body_intrusion: -0.2608 Episode_Reward/pair_body_intrusion: -0.3237 Curriculum/terrain_level_mean: +3.9084 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.5067 Episode_Termination/fell_over: 0.0267 Episode_Termination/course_exit: 0.4400 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[823] RECOVERY reset 1 invalid environments; first IDs=[2734] ============================================================================== Learning iteration 31350/50000 Total steps: 128,409,600 Steps per second: 13,084 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1411.0 Actor loss: -9.5431 Critic loss: +0.0051 World model loss: +0.6423 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1628 Episode_Reward/track_angular_velocity: +53.4117 Episode_Reward/upright: +27.0782 Episode_Reward/leg_posture: -2.7334 Episode_Reward/standstill: -6.8337 Episode_Reward/action_rate_l2: -0.5814 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0083 Episode_Reward/unnecessary_wheel_lift: -0.0872 Episode_Reward/rolling_leg_motion: -0.6314 Episode_Reward/obstacle_progress: +3.7232 Episode_Reward/wheel_body_intrusion: -0.3068 Episode_Reward/pair_body_intrusion: -0.3578 Curriculum/terrain_level_mean: +3.9128 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4676 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5108 Episode_Termination/illegal_nonwheel_contact: 0.0216 ============================================================================== ============================================================================== Learning iteration 31400/50000 Total steps: 128,614,400 Steps per second: 13,088 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1317.8 Actor loss: -9.5500 Critic loss: +0.0034 World model loss: +0.6141 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8151 Episode_Reward/track_angular_velocity: +49.8697 Episode_Reward/upright: +25.2624 Episode_Reward/leg_posture: -2.5374 Episode_Reward/standstill: -6.1450 Episode_Reward/action_rate_l2: -0.5439 Episode_Reward/joint_torques_l2: -0.1409 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0154 Episode_Reward/unnecessary_wheel_lift: -0.0822 Episode_Reward/rolling_leg_motion: -0.5716 Episode_Reward/obstacle_progress: +3.5043 Episode_Reward/wheel_body_intrusion: -0.2699 Episode_Reward/pair_body_intrusion: -0.3243 Curriculum/terrain_level_mean: +3.9150 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4097 Episode_Termination/fell_over: 0.0208 Episode_Termination/course_exit: 0.5486 Episode_Termination/illegal_nonwheel_contact: 0.0347 ============================================================================== ============================================================================== Learning iteration 31450/50000 Total steps: 128,819,200 Steps per second: 13,091 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1400.7 Actor loss: -9.5081 Critic loss: +0.0050 World model loss: +0.5804 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7118 Episode_Reward/track_angular_velocity: +53.0105 Episode_Reward/upright: +26.8575 Episode_Reward/leg_posture: -2.6854 Episode_Reward/standstill: -6.5150 Episode_Reward/action_rate_l2: -0.5849 Episode_Reward/joint_torques_l2: -0.1505 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0975 Episode_Reward/rolling_leg_motion: -0.6263 Episode_Reward/obstacle_progress: +3.7550 Episode_Reward/wheel_body_intrusion: -0.2652 Episode_Reward/pair_body_intrusion: -0.2916 Curriculum/terrain_level_mean: +3.9172 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.4859 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.4859 Episode_Termination/illegal_nonwheel_contact: 0.0141 ============================================================================== ============================================================================== Learning iteration 31500/50000 Total steps: 129,024,000 Steps per second: 13,095 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1326.4 Actor loss: -9.5178 Critic loss: +0.0042 World model loss: +0.6032 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9041 Episode_Reward/track_angular_velocity: +50.0851 Episode_Reward/upright: +25.4308 Episode_Reward/leg_posture: -2.5430 Episode_Reward/standstill: -6.3363 Episode_Reward/action_rate_l2: -0.5639 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0389 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.5927 Episode_Reward/obstacle_progress: +3.6107 Episode_Reward/wheel_body_intrusion: -0.2678 Episode_Reward/pair_body_intrusion: -0.2770 Curriculum/terrain_level_mean: +3.9172 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4503 Episode_Termination/fell_over: 0.0596 Episode_Termination/course_exit: 0.4503 Episode_Termination/illegal_nonwheel_contact: 0.0530 ============================================================================== ============================================================================== Learning iteration 31550/50000 Total steps: 129,228,800 Steps per second: 13,098 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0871 Mean episode length: 1349.7 Actor loss: -9.5321 Critic loss: +0.0029 World model loss: +0.6452 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7327 Episode_Reward/track_angular_velocity: +51.0108 Episode_Reward/upright: +25.9063 Episode_Reward/leg_posture: -2.5588 Episode_Reward/standstill: -6.1956 Episode_Reward/action_rate_l2: -0.5566 Episode_Reward/joint_torques_l2: -0.1446 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0922 Episode_Reward/rolling_leg_motion: -0.5874 Episode_Reward/obstacle_progress: +3.6411 Episode_Reward/wheel_body_intrusion: -0.2690 Episode_Reward/pair_body_intrusion: -0.2929 Curriculum/terrain_level_mean: +3.9189 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5098 Episode_Termination/fell_over: 0.0131 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 31600/50000 Total steps: 129,433,600 Steps per second: 13,101 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1318.0 Actor loss: -9.5168 Critic loss: +0.0042 World model loss: +0.6006 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9410 Episode_Reward/track_angular_velocity: +49.8826 Episode_Reward/upright: +25.2955 Episode_Reward/leg_posture: -2.4843 Episode_Reward/standstill: -5.9301 Episode_Reward/action_rate_l2: -0.5474 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0727 Episode_Reward/rolling_leg_motion: -0.5623 Episode_Reward/obstacle_progress: +3.6329 Episode_Reward/wheel_body_intrusion: -0.2279 Episode_Reward/pair_body_intrusion: -0.2742 Curriculum/terrain_level_mean: +3.9197 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4453 Episode_Termination/fell_over: 0.0234 Episode_Termination/course_exit: 0.4922 Episode_Termination/illegal_nonwheel_contact: 0.0547 ============================================================================== ============================================================================== Learning iteration 31650/50000 Total steps: 129,638,400 Steps per second: 13,104 Iteration time: 0.313s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1389.1 Actor loss: -9.4850 Critic loss: +0.0085 World model loss: +0.6102 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4465 Episode_Reward/track_angular_velocity: +52.5935 Episode_Reward/upright: +26.6716 Episode_Reward/leg_posture: -2.6342 Episode_Reward/standstill: -6.3351 Episode_Reward/action_rate_l2: -0.5884 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.0987 Episode_Reward/rolling_leg_motion: -0.6012 Episode_Reward/obstacle_progress: +3.7981 Episode_Reward/wheel_body_intrusion: -0.2441 Episode_Reward/pair_body_intrusion: -0.2776 Curriculum/terrain_level_mean: +3.9207 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.4771 Episode_Termination/fell_over: 0.0327 Episode_Termination/course_exit: 0.4575 Episode_Termination/illegal_nonwheel_contact: 0.0588 ============================================================================== ============================================================================== Learning iteration 31700/50000 Total steps: 129,843,200 Steps per second: 13,108 Iteration time: 0.312s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1429.3 Actor loss: -9.5126 Critic loss: +0.0051 World model loss: +0.5926 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.0199 Episode_Reward/track_angular_velocity: +54.0982 Episode_Reward/upright: +27.4694 Episode_Reward/leg_posture: -2.7263 Episode_Reward/standstill: -6.8350 Episode_Reward/action_rate_l2: -0.6027 Episode_Reward/joint_torques_l2: -0.1526 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0112 Episode_Reward/unnecessary_wheel_lift: -0.0893 Episode_Reward/rolling_leg_motion: -0.6365 Episode_Reward/obstacle_progress: +3.9291 Episode_Reward/wheel_body_intrusion: -0.2549 Episode_Reward/pair_body_intrusion: -0.2755 Curriculum/terrain_level_mean: +3.9231 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0903 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.5714 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.4014 Episode_Termination/illegal_nonwheel_contact: 0.0204 ============================================================================== ============================================================================== Learning iteration 31750/50000 Total steps: 130,048,000 Steps per second: 13,111 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1372.4 Actor loss: -9.4981 Critic loss: +0.0047 World model loss: +0.6084 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8915 Episode_Reward/track_angular_velocity: +51.9633 Episode_Reward/upright: +26.3634 Episode_Reward/leg_posture: -2.5626 Episode_Reward/standstill: -6.3328 Episode_Reward/action_rate_l2: -0.5781 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.0887 Episode_Reward/rolling_leg_motion: -0.6235 Episode_Reward/obstacle_progress: +3.7451 Episode_Reward/wheel_body_intrusion: -0.2601 Episode_Reward/pair_body_intrusion: -0.3038 Curriculum/terrain_level_mean: +3.9253 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.4516 Episode_Termination/fell_over: 0.0258 Episode_Termination/course_exit: 0.4968 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1007] ============================================================================== Learning iteration 31800/50000 Total steps: 130,252,800 Steps per second: 13,115 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1357.9 Actor loss: -9.4914 Critic loss: +0.0076 World model loss: +0.6154 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4392 Episode_Reward/track_angular_velocity: +51.4478 Episode_Reward/upright: +26.1146 Episode_Reward/leg_posture: -2.5221 Episode_Reward/standstill: -6.2586 Episode_Reward/action_rate_l2: -0.5736 Episode_Reward/joint_torques_l2: -0.1465 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0053 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.6061 Episode_Reward/obstacle_progress: +3.7271 Episode_Reward/wheel_body_intrusion: -0.2429 Episode_Reward/pair_body_intrusion: -0.2341 Curriculum/terrain_level_mean: +3.9297 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1314 Episode_Termination/time_out: 0.4933 Episode_Termination/fell_over: 0.0067 Episode_Termination/course_exit: 0.4933 Episode_Termination/illegal_nonwheel_contact: 0.0133 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1388] ============================================================================== Learning iteration 31850/50000 Total steps: 130,457,600 Steps per second: 13,118 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1373.9 Actor loss: -9.4912 Critic loss: +0.0059 World model loss: +0.6147 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1775 Episode_Reward/track_angular_velocity: +52.0553 Episode_Reward/upright: +26.3983 Episode_Reward/leg_posture: -2.6085 Episode_Reward/standstill: -6.2684 Episode_Reward/action_rate_l2: -0.5751 Episode_Reward/joint_torques_l2: -0.1471 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.6044 Episode_Reward/obstacle_progress: +3.8171 Episode_Reward/wheel_body_intrusion: -0.2444 Episode_Reward/pair_body_intrusion: -0.2523 Curriculum/terrain_level_mean: +3.9326 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0889 Metrics/base_velocity/error_vel_yaw: 0.1316 Episode_Termination/time_out: 0.4774 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4903 Episode_Termination/illegal_nonwheel_contact: 0.0129 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1873] RECOVERY reset 1 invalid environments; first IDs=[2412] ============================================================================== Learning iteration 31900/50000 Total steps: 130,662,400 Steps per second: 13,121 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1385.8 Actor loss: -9.5403 Critic loss: +0.0034 World model loss: +0.6089 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3824 Episode_Reward/track_angular_velocity: +52.5238 Episode_Reward/upright: +26.6291 Episode_Reward/leg_posture: -2.6755 Episode_Reward/standstill: -6.6325 Episode_Reward/action_rate_l2: -0.5817 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0849 Episode_Reward/rolling_leg_motion: -0.6127 Episode_Reward/obstacle_progress: +3.6864 Episode_Reward/wheel_body_intrusion: -0.2590 Episode_Reward/pair_body_intrusion: -0.3062 Curriculum/terrain_level_mean: +3.9355 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1311 Episode_Termination/time_out: 0.4702 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.5033 Episode_Termination/illegal_nonwheel_contact: 0.0199 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1394] ============================================================================== Learning iteration 31950/50000 Total steps: 130,867,200 Steps per second: 13,125 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0873 Mean episode length: 1416.9 Actor loss: -9.5363 Critic loss: +0.0035 World model loss: +0.6246 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3464 Episode_Reward/track_angular_velocity: +53.6767 Episode_Reward/upright: +27.2188 Episode_Reward/leg_posture: -2.7283 Episode_Reward/standstill: -6.5079 Episode_Reward/action_rate_l2: -0.6033 Episode_Reward/joint_torques_l2: -0.1549 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1023 Episode_Reward/rolling_leg_motion: -0.6278 Episode_Reward/obstacle_progress: +3.7613 Episode_Reward/wheel_body_intrusion: -0.2921 Episode_Reward/pair_body_intrusion: -0.3344 Curriculum/terrain_level_mean: +3.9385 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1318 Episode_Termination/time_out: 0.5325 Episode_Termination/fell_over: 0.0296 Episode_Termination/course_exit: 0.4201 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1416] RECOVERY reset 1 invalid environments; first IDs=[1043] ============================================================================== Learning iteration 32000/50000 Total steps: 131,072,000 Steps per second: 13,128 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1344.9 Actor loss: -9.5181 Critic loss: +0.0044 World model loss: +0.5721 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8053 Episode_Reward/track_angular_velocity: +50.9471 Episode_Reward/upright: +25.8553 Episode_Reward/leg_posture: -2.5621 Episode_Reward/standstill: -6.3232 Episode_Reward/action_rate_l2: -0.5595 Episode_Reward/joint_torques_l2: -0.1465 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0840 Episode_Reward/rolling_leg_motion: -0.5872 Episode_Reward/obstacle_progress: +3.5938 Episode_Reward/wheel_body_intrusion: -0.2507 Episode_Reward/pair_body_intrusion: -0.2622 Curriculum/terrain_level_mean: +3.9404 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.4638 Episode_Termination/fell_over: 0.0145 Episode_Termination/course_exit: 0.5072 Episode_Termination/illegal_nonwheel_contact: 0.0217 ============================================================================== ============================================================================== Learning iteration 32050/50000 Total steps: 131,276,800 Steps per second: 13,131 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1311.9 Actor loss: -9.5569 Critic loss: +0.0029 World model loss: +0.5927 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6333 Episode_Reward/track_angular_velocity: +49.6655 Episode_Reward/upright: +25.1983 Episode_Reward/leg_posture: -2.5339 Episode_Reward/standstill: -6.1329 Episode_Reward/action_rate_l2: -0.5560 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0179 Episode_Reward/unnecessary_wheel_lift: -0.0804 Episode_Reward/rolling_leg_motion: -0.5969 Episode_Reward/obstacle_progress: +3.5058 Episode_Reward/wheel_body_intrusion: -0.2606 Episode_Reward/pair_body_intrusion: -0.2475 Curriculum/terrain_level_mean: +3.9424 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.4296 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.5211 Episode_Termination/illegal_nonwheel_contact: 0.0423 ============================================================================== ============================================================================== Learning iteration 32100/50000 Total steps: 131,481,600 Steps per second: 13,135 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1410.1 Actor loss: -9.5353 Critic loss: +0.0062 World model loss: +0.6189 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9854 Episode_Reward/track_angular_velocity: +53.3118 Episode_Reward/upright: +27.0720 Episode_Reward/leg_posture: -2.6826 Episode_Reward/standstill: -6.5585 Episode_Reward/action_rate_l2: -0.5939 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0248 Episode_Reward/unnecessary_wheel_lift: -0.0917 Episode_Reward/rolling_leg_motion: -0.6239 Episode_Reward/obstacle_progress: +3.9216 Episode_Reward/wheel_body_intrusion: -0.2738 Episode_Reward/pair_body_intrusion: -0.3042 Curriculum/terrain_level_mean: +3.9429 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5395 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.4013 Episode_Termination/illegal_nonwheel_contact: 0.0461 ============================================================================== ============================================================================== Learning iteration 32150/50000 Total steps: 131,686,400 Steps per second: 13,138 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1324.7 Actor loss: -9.5703 Critic loss: +0.0058 World model loss: +0.5987 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9671 Episode_Reward/track_angular_velocity: +50.0899 Episode_Reward/upright: +25.4580 Episode_Reward/leg_posture: -2.5591 Episode_Reward/standstill: -5.8135 Episode_Reward/action_rate_l2: -0.5559 Episode_Reward/joint_torques_l2: -0.1443 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0233 Episode_Reward/unnecessary_wheel_lift: -0.0976 Episode_Reward/rolling_leg_motion: -0.5935 Episode_Reward/obstacle_progress: +3.6156 Episode_Reward/wheel_body_intrusion: -0.2621 Episode_Reward/pair_body_intrusion: -0.2820 Curriculum/terrain_level_mean: +3.9438 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4788 Episode_Termination/fell_over: 0.0303 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0424 ============================================================================== ============================================================================== Learning iteration 32200/50000 Total steps: 131,891,200 Steps per second: 13,141 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1374.6 Actor loss: -9.5399 Critic loss: +0.0037 World model loss: +0.5606 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9287 Episode_Reward/track_angular_velocity: +51.9580 Episode_Reward/upright: +26.4028 Episode_Reward/leg_posture: -2.6583 Episode_Reward/standstill: -5.9951 Episode_Reward/action_rate_l2: -0.5832 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0898 Episode_Reward/rolling_leg_motion: -0.6139 Episode_Reward/obstacle_progress: +3.8263 Episode_Reward/wheel_body_intrusion: -0.2685 Episode_Reward/pair_body_intrusion: -0.2843 Curriculum/terrain_level_mean: +3.9470 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0207 Episode_Termination/course_exit: 0.4483 Episode_Termination/illegal_nonwheel_contact: 0.0276 ============================================================================== ============================================================================== Learning iteration 32250/50000 Total steps: 132,096,000 Steps per second: 13,145 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1379.9 Actor loss: -9.5148 Critic loss: +0.0078 World model loss: +0.6053 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1594 Episode_Reward/track_angular_velocity: +52.2013 Episode_Reward/upright: +26.5383 Episode_Reward/leg_posture: -2.6481 Episode_Reward/standstill: -6.0561 Episode_Reward/action_rate_l2: -0.5920 Episode_Reward/joint_torques_l2: -0.1498 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0270 Episode_Reward/unnecessary_wheel_lift: -0.0979 Episode_Reward/rolling_leg_motion: -0.6252 Episode_Reward/obstacle_progress: +3.9256 Episode_Reward/wheel_body_intrusion: -0.2502 Episode_Reward/pair_body_intrusion: -0.3135 Curriculum/terrain_level_mean: +3.9475 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0429 Episode_Termination/course_exit: 0.4417 Episode_Termination/illegal_nonwheel_contact: 0.0307 ============================================================================== ============================================================================== Learning iteration 32300/50000 Total steps: 132,300,800 Steps per second: 13,148 Iteration time: 0.312s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0881 Mean episode length: 1361.3 Actor loss: -9.5482 Critic loss: +0.0033 World model loss: +0.6380 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5807 Episode_Reward/track_angular_velocity: +51.5658 Episode_Reward/upright: +26.1904 Episode_Reward/leg_posture: -2.5805 Episode_Reward/standstill: -5.6842 Episode_Reward/action_rate_l2: -0.5693 Episode_Reward/joint_torques_l2: -0.1470 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0773 Episode_Reward/rolling_leg_motion: -0.5975 Episode_Reward/obstacle_progress: +3.8557 Episode_Reward/wheel_body_intrusion: -0.2565 Episode_Reward/pair_body_intrusion: -0.3278 Curriculum/terrain_level_mean: +3.9490 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0901 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.4196 Episode_Termination/fell_over: 0.0280 Episode_Termination/course_exit: 0.5385 Episode_Termination/illegal_nonwheel_contact: 0.0280 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[548] RECOVERY reset 1 invalid environments; first IDs=[1303] ============================================================================== Learning iteration 32350/50000 Total steps: 132,505,600 Steps per second: 13,151 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0878 Mean episode length: 1305.4 Actor loss: -9.5417 Critic loss: +0.0039 World model loss: +0.6423 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5370 Episode_Reward/track_angular_velocity: +49.4068 Episode_Reward/upright: +25.0868 Episode_Reward/leg_posture: -2.5036 Episode_Reward/standstill: -5.6489 Episode_Reward/action_rate_l2: -0.5459 Episode_Reward/joint_torques_l2: -0.1412 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0229 Episode_Reward/unnecessary_wheel_lift: -0.0882 Episode_Reward/rolling_leg_motion: -0.5856 Episode_Reward/obstacle_progress: +3.6034 Episode_Reward/wheel_body_intrusion: -0.2417 Episode_Reward/pair_body_intrusion: -0.2706 Curriculum/terrain_level_mean: +3.9507 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4048 Episode_Termination/fell_over: 0.0238 Episode_Termination/course_exit: 0.5417 Episode_Termination/illegal_nonwheel_contact: 0.0357 ============================================================================== ============================================================================== Learning iteration 32400/50000 Total steps: 132,710,400 Steps per second: 13,154 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1325.1 Actor loss: -9.5039 Critic loss: +0.0032 World model loss: +0.6053 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0486 Episode_Reward/track_angular_velocity: +50.1578 Episode_Reward/upright: +25.4353 Episode_Reward/leg_posture: -2.5036 Episode_Reward/standstill: -5.7175 Episode_Reward/action_rate_l2: -0.5574 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0285 Episode_Reward/unnecessary_wheel_lift: -0.0872 Episode_Reward/rolling_leg_motion: -0.5941 Episode_Reward/obstacle_progress: +3.6203 Episode_Reward/wheel_body_intrusion: -0.2421 Episode_Reward/pair_body_intrusion: -0.2773 Curriculum/terrain_level_mean: +3.9529 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.4333 Episode_Termination/fell_over: 0.0333 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0467 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_032400.mp4 ============================================================================== Learning iteration 32450/50000 Total steps: 132,915,200 Steps per second: 13,150 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1370.5 Actor loss: -9.5036 Critic loss: +0.0046 World model loss: +0.6402 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7102 Episode_Reward/track_angular_velocity: +51.8438 Episode_Reward/upright: +26.3159 Episode_Reward/leg_posture: -2.5812 Episode_Reward/standstill: -5.9890 Episode_Reward/action_rate_l2: -0.5763 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0842 Episode_Reward/rolling_leg_motion: -0.5948 Episode_Reward/obstacle_progress: +3.8108 Episode_Reward/wheel_body_intrusion: -0.2592 Episode_Reward/pair_body_intrusion: -0.3041 Curriculum/terrain_level_mean: +3.9556 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4551 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.5064 Episode_Termination/illegal_nonwheel_contact: 0.0256 ============================================================================== ============================================================================== Learning iteration 32500/50000 Total steps: 133,120,000 Steps per second: 13,153 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1389.8 Actor loss: -9.5328 Critic loss: +0.0067 World model loss: +0.5902 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4768 Episode_Reward/track_angular_velocity: +52.6078 Episode_Reward/upright: +26.6684 Episode_Reward/leg_posture: -2.6358 Episode_Reward/standstill: -6.0811 Episode_Reward/action_rate_l2: -0.5889 Episode_Reward/joint_torques_l2: -0.1509 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0271 Episode_Reward/unnecessary_wheel_lift: -0.0798 Episode_Reward/rolling_leg_motion: -0.5841 Episode_Reward/obstacle_progress: +3.8409 Episode_Reward/wheel_body_intrusion: -0.2722 Episode_Reward/pair_body_intrusion: -0.3025 Curriculum/terrain_level_mean: +3.9573 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.5130 Episode_Termination/fell_over: 0.0649 Episode_Termination/course_exit: 0.4221 Episode_Termination/illegal_nonwheel_contact: 0.0195 ============================================================================== ============================================================================== Learning iteration 32550/50000 Total steps: 133,324,800 Steps per second: 13,157 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0879 Mean episode length: 1443.5 Actor loss: -9.5375 Critic loss: +0.0040 World model loss: +0.6353 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.4224 Episode_Reward/track_angular_velocity: +54.6803 Episode_Reward/upright: +27.7273 Episode_Reward/leg_posture: -2.7471 Episode_Reward/standstill: -6.0800 Episode_Reward/action_rate_l2: -0.6123 Episode_Reward/joint_torques_l2: -0.1547 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0113 Episode_Reward/unnecessary_wheel_lift: -0.0834 Episode_Reward/rolling_leg_motion: -0.6180 Episode_Reward/obstacle_progress: +3.9548 Episode_Reward/wheel_body_intrusion: -0.2700 Episode_Reward/pair_body_intrusion: -0.2801 Curriculum/terrain_level_mean: +3.9590 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0907 Metrics/base_velocity/error_vel_yaw: 0.1321 Episode_Termination/time_out: 0.5548 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4178 Episode_Termination/illegal_nonwheel_contact: 0.0274 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[493] ============================================================================== Learning iteration 32600/50000 Total steps: 133,529,600 Steps per second: 13,160 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0878 Mean episode length: 1417.0 Actor loss: -9.5483 Critic loss: +0.0066 World model loss: +0.6264 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3512 Episode_Reward/track_angular_velocity: +53.6052 Episode_Reward/upright: +27.2361 Episode_Reward/leg_posture: -2.6689 Episode_Reward/standstill: -5.9053 Episode_Reward/action_rate_l2: -0.6188 Episode_Reward/joint_torques_l2: -0.1536 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.0824 Episode_Reward/rolling_leg_motion: -0.6198 Episode_Reward/obstacle_progress: +3.9252 Episode_Reward/wheel_body_intrusion: -0.2606 Episode_Reward/pair_body_intrusion: -0.3076 Curriculum/terrain_level_mean: +3.9612 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5578 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4150 Episode_Termination/illegal_nonwheel_contact: 0.0136 ============================================================================== ============================================================================== Learning iteration 32650/50000 Total steps: 133,734,400 Steps per second: 13,163 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0877 Mean episode length: 1341.5 Actor loss: -9.5083 Critic loss: +0.0049 World model loss: +0.6398 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7430 Episode_Reward/track_angular_velocity: +50.7964 Episode_Reward/upright: +25.7742 Episode_Reward/leg_posture: -2.5227 Episode_Reward/standstill: -5.8727 Episode_Reward/action_rate_l2: -0.5661 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0881 Episode_Reward/rolling_leg_motion: -0.5852 Episode_Reward/obstacle_progress: +3.7004 Episode_Reward/wheel_body_intrusion: -0.2367 Episode_Reward/pair_body_intrusion: -0.2823 Curriculum/terrain_level_mean: +3.9641 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1321 Episode_Termination/time_out: 0.4052 Episode_Termination/fell_over: 0.0196 Episode_Termination/course_exit: 0.5556 Episode_Termination/illegal_nonwheel_contact: 0.0261 ============================================================================== ============================================================================== Learning iteration 32700/50000 Total steps: 133,939,200 Steps per second: 13,167 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1432.4 Actor loss: -9.5299 Critic loss: +0.0064 World model loss: +0.6235 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.9544 Episode_Reward/track_angular_velocity: +54.2175 Episode_Reward/upright: +27.4852 Episode_Reward/leg_posture: -2.7490 Episode_Reward/standstill: -6.4281 Episode_Reward/action_rate_l2: -0.6019 Episode_Reward/joint_torques_l2: -0.1544 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0112 Episode_Reward/unnecessary_wheel_lift: -0.0855 Episode_Reward/rolling_leg_motion: -0.6176 Episode_Reward/obstacle_progress: +3.8764 Episode_Reward/wheel_body_intrusion: -0.2672 Episode_Reward/pair_body_intrusion: -0.3098 Curriculum/terrain_level_mean: +3.9666 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.5512 Episode_Termination/fell_over: 0.0315 Episode_Termination/course_exit: 0.4173 Episode_Termination/illegal_nonwheel_contact: 0.0079 ============================================================================== ============================================================================== Learning iteration 32750/50000 Total steps: 134,144,000 Steps per second: 13,170 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1360.8 Actor loss: -9.5317 Critic loss: +0.0055 World model loss: +0.5482 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3373 Episode_Reward/track_angular_velocity: +51.5385 Episode_Reward/upright: +26.1243 Episode_Reward/leg_posture: -2.5993 Episode_Reward/standstill: -5.7727 Episode_Reward/action_rate_l2: -0.5687 Episode_Reward/joint_torques_l2: -0.1475 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0266 Episode_Reward/unnecessary_wheel_lift: -0.0869 Episode_Reward/rolling_leg_motion: -0.6003 Episode_Reward/obstacle_progress: +3.6208 Episode_Reward/wheel_body_intrusion: -0.2648 Episode_Reward/pair_body_intrusion: -0.3133 Curriculum/terrain_level_mean: +3.9675 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.4345 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.5060 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 32800/50000 Total steps: 134,348,800 Steps per second: 13,173 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0878 Mean episode length: 1406.7 Actor loss: -9.5520 Critic loss: +0.0032 World model loss: +0.5829 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1170 Episode_Reward/track_angular_velocity: +53.2711 Episode_Reward/upright: +27.0141 Episode_Reward/leg_posture: -2.6479 Episode_Reward/standstill: -5.9922 Episode_Reward/action_rate_l2: -0.6058 Episode_Reward/joint_torques_l2: -0.1518 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.0715 Episode_Reward/rolling_leg_motion: -0.6189 Episode_Reward/obstacle_progress: +3.7836 Episode_Reward/wheel_body_intrusion: -0.2635 Episode_Reward/pair_body_intrusion: -0.2911 Curriculum/terrain_level_mean: +3.9685 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.5177 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.4397 Episode_Termination/illegal_nonwheel_contact: 0.0213 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[551] ============================================================================== Learning iteration 32850/50000 Total steps: 134,553,600 Steps per second: 13,177 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1335.0 Actor loss: -9.5199 Critic loss: +0.0047 World model loss: +0.5909 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5751 Episode_Reward/track_angular_velocity: +50.4886 Episode_Reward/upright: +25.6323 Episode_Reward/leg_posture: -2.5205 Episode_Reward/standstill: -5.8118 Episode_Reward/action_rate_l2: -0.5635 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.0864 Episode_Reward/rolling_leg_motion: -0.5851 Episode_Reward/obstacle_progress: +3.6150 Episode_Reward/wheel_body_intrusion: -0.2400 Episode_Reward/pair_body_intrusion: -0.2760 Curriculum/terrain_level_mean: +3.9697 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0908 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4178 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.5479 Episode_Termination/illegal_nonwheel_contact: 0.0274 ============================================================================== ============================================================================== Learning iteration 32900/50000 Total steps: 134,758,400 Steps per second: 13,180 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1275.6 Actor loss: -9.5340 Critic loss: +0.0035 World model loss: +0.5791 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.2334 Episode_Reward/track_angular_velocity: +48.2964 Episode_Reward/upright: +24.4889 Episode_Reward/leg_posture: -2.4163 Episode_Reward/standstill: -5.7952 Episode_Reward/action_rate_l2: -0.5458 Episode_Reward/joint_torques_l2: -0.1372 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0224 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5658 Episode_Reward/obstacle_progress: +3.4797 Episode_Reward/wheel_body_intrusion: -0.2363 Episode_Reward/pair_body_intrusion: -0.2641 Curriculum/terrain_level_mean: +3.9712 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.4296 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.5070 Episode_Termination/illegal_nonwheel_contact: 0.0634 ============================================================================== ============================================================================== Learning iteration 32950/50000 Total steps: 134,963,200 Steps per second: 13,183 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1334.8 Actor loss: -9.5296 Critic loss: +0.0051 World model loss: +0.6168 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2938 Episode_Reward/track_angular_velocity: +50.5293 Episode_Reward/upright: +25.6265 Episode_Reward/leg_posture: -2.5421 Episode_Reward/standstill: -5.7560 Episode_Reward/action_rate_l2: -0.5657 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0104 Episode_Reward/unnecessary_wheel_lift: -0.0929 Episode_Reward/rolling_leg_motion: -0.5886 Episode_Reward/obstacle_progress: +3.7292 Episode_Reward/wheel_body_intrusion: -0.2645 Episode_Reward/pair_body_intrusion: -0.2861 Curriculum/terrain_level_mean: +3.9731 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4241 Episode_Termination/fell_over: 0.0127 Episode_Termination/course_exit: 0.5506 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 33000/50000 Total steps: 135,168,000 Steps per second: 13,186 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0879 Mean episode length: 1296.4 Actor loss: -9.5531 Critic loss: +0.0034 World model loss: +0.6042 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1136 Episode_Reward/track_angular_velocity: +49.0879 Episode_Reward/upright: +24.8885 Episode_Reward/leg_posture: -2.4564 Episode_Reward/standstill: -5.4057 Episode_Reward/action_rate_l2: -0.5489 Episode_Reward/joint_torques_l2: -0.1395 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0253 Episode_Reward/unnecessary_wheel_lift: -0.0829 Episode_Reward/rolling_leg_motion: -0.5735 Episode_Reward/obstacle_progress: +3.6133 Episode_Reward/wheel_body_intrusion: -0.2460 Episode_Reward/pair_body_intrusion: -0.2425 Curriculum/terrain_level_mean: +3.9746 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.4167 Episode_Termination/fell_over: 0.0385 Episode_Termination/course_exit: 0.5192 Episode_Termination/illegal_nonwheel_contact: 0.0385 ============================================================================== ============================================================================== Learning iteration 33050/50000 Total steps: 135,372,800 Steps per second: 13,189 Iteration time: 0.311s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0880 Mean episode length: 1332.9 Actor loss: -9.5142 Critic loss: +0.0138 World model loss: +0.5786 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3775 Episode_Reward/track_angular_velocity: +50.4538 Episode_Reward/upright: +25.5720 Episode_Reward/leg_posture: -2.4787 Episode_Reward/standstill: -5.3522 Episode_Reward/action_rate_l2: -0.5717 Episode_Reward/joint_torques_l2: -0.1431 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0078 Episode_Reward/unnecessary_wheel_lift: -0.0831 Episode_Reward/rolling_leg_motion: -0.5946 Episode_Reward/obstacle_progress: +3.6590 Episode_Reward/wheel_body_intrusion: -0.2307 Episode_Reward/pair_body_intrusion: -0.2622 Curriculum/terrain_level_mean: +3.9785 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.4118 Episode_Termination/fell_over: 0.0059 Episode_Termination/course_exit: 0.5706 Episode_Termination/illegal_nonwheel_contact: 0.0118 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2176] RECOVERY reset 1 invalid environments; first IDs=[953] ============================================================================== Learning iteration 33100/50000 Total steps: 135,577,600 Steps per second: 13,192 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0880 Mean episode length: 1307.3 Actor loss: -9.4814 Critic loss: +0.0065 World model loss: +0.5742 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4040 Episode_Reward/track_angular_velocity: +49.4779 Episode_Reward/upright: +25.0572 Episode_Reward/leg_posture: -2.4183 Episode_Reward/standstill: -5.2375 Episode_Reward/action_rate_l2: -0.5626 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0281 Episode_Reward/unnecessary_wheel_lift: -0.0764 Episode_Reward/rolling_leg_motion: -0.5758 Episode_Reward/obstacle_progress: +3.6740 Episode_Reward/wheel_body_intrusion: -0.2368 Episode_Reward/pair_body_intrusion: -0.2554 Curriculum/terrain_level_mean: +3.9785 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4228 Episode_Termination/fell_over: 0.0268 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0537 ============================================================================== ============================================================================== Learning iteration 33150/50000 Total steps: 135,782,400 Steps per second: 13,195 Iteration time: 0.310s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0878 Mean episode length: 1334.8 Actor loss: -9.5208 Critic loss: +0.0037 World model loss: +0.6583 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3571 Episode_Reward/track_angular_velocity: +50.5507 Episode_Reward/upright: +25.5734 Episode_Reward/leg_posture: -2.5122 Episode_Reward/standstill: -5.4805 Episode_Reward/action_rate_l2: -0.5740 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.0789 Episode_Reward/rolling_leg_motion: -0.5797 Episode_Reward/obstacle_progress: +3.7050 Episode_Reward/wheel_body_intrusion: -0.2533 Episode_Reward/pair_body_intrusion: -0.2905 Curriculum/terrain_level_mean: +3.9807 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.4403 Episode_Termination/fell_over: 0.0224 Episode_Termination/course_exit: 0.5149 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 33200/50000 Total steps: 135,987,200 Steps per second: 13,199 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0877 Mean episode length: 1383.3 Actor loss: -9.5375 Critic loss: +0.0054 World model loss: +0.6141 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0146 Episode_Reward/track_angular_velocity: +52.3387 Episode_Reward/upright: +26.4621 Episode_Reward/leg_posture: -2.5561 Episode_Reward/standstill: -5.9024 Episode_Reward/action_rate_l2: -0.5906 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0270 Episode_Reward/unnecessary_wheel_lift: -0.0719 Episode_Reward/rolling_leg_motion: -0.5768 Episode_Reward/obstacle_progress: +3.9032 Episode_Reward/wheel_body_intrusion: -0.2638 Episode_Reward/pair_body_intrusion: -0.2798 Curriculum/terrain_level_mean: +3.9819 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4621 Episode_Termination/fell_over: 0.0227 Episode_Termination/course_exit: 0.4621 Episode_Termination/illegal_nonwheel_contact: 0.0530 ============================================================================== ============================================================================== Learning iteration 33250/50000 Total steps: 136,192,000 Steps per second: 13,202 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0877 Mean episode length: 1431.6 Actor loss: -9.5135 Critic loss: +0.0060 World model loss: +0.6203 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7560 Episode_Reward/track_angular_velocity: +54.1604 Episode_Reward/upright: +27.4102 Episode_Reward/leg_posture: -2.6918 Episode_Reward/standstill: -6.0386 Episode_Reward/action_rate_l2: -0.6086 Episode_Reward/joint_torques_l2: -0.1552 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0112 Episode_Reward/unnecessary_wheel_lift: -0.0766 Episode_Reward/rolling_leg_motion: -0.6214 Episode_Reward/obstacle_progress: +4.0203 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.3194 Curriculum/terrain_level_mean: +3.9844 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5655 Episode_Termination/fell_over: 0.0207 Episode_Termination/course_exit: 0.4069 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 33300/50000 Total steps: 136,396,800 Steps per second: 13,205 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0873 Mean episode length: 1369.9 Actor loss: -9.5185 Critic loss: +0.0062 World model loss: +0.5949 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5406 Episode_Reward/track_angular_velocity: +51.8717 Episode_Reward/upright: +26.2477 Episode_Reward/leg_posture: -2.5668 Episode_Reward/standstill: -6.1850 Episode_Reward/action_rate_l2: -0.5689 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0268 Episode_Reward/unnecessary_wheel_lift: -0.1062 Episode_Reward/rolling_leg_motion: -0.6001 Episode_Reward/obstacle_progress: +3.7317 Episode_Reward/wheel_body_intrusion: -0.2673 Episode_Reward/pair_body_intrusion: -0.3044 Curriculum/terrain_level_mean: +3.9858 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4662 Episode_Termination/fell_over: 0.0541 Episode_Termination/course_exit: 0.4662 Episode_Termination/illegal_nonwheel_contact: 0.0405 ============================================================================== ============================================================================== Learning iteration 33350/50000 Total steps: 136,601,600 Steps per second: 13,209 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0876 Mean episode length: 1304.4 Actor loss: -9.5062 Critic loss: +0.0054 World model loss: +0.6138 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2593 Episode_Reward/track_angular_velocity: +49.3874 Episode_Reward/upright: +25.0353 Episode_Reward/leg_posture: -2.4553 Episode_Reward/standstill: -5.6920 Episode_Reward/action_rate_l2: -0.5289 Episode_Reward/joint_torques_l2: -0.1402 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0102 Episode_Reward/unnecessary_wheel_lift: -0.0795 Episode_Reward/rolling_leg_motion: -0.5564 Episode_Reward/obstacle_progress: +3.5417 Episode_Reward/wheel_body_intrusion: -0.2433 Episode_Reward/pair_body_intrusion: -0.2854 Curriculum/terrain_level_mean: +3.9888 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4000 Episode_Termination/fell_over: 0.0069 Episode_Termination/course_exit: 0.5724 Episode_Termination/illegal_nonwheel_contact: 0.0276 ============================================================================== ============================================================================== Learning iteration 33400/50000 Total steps: 136,806,400 Steps per second: 13,212 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1314.6 Actor loss: -9.5425 Critic loss: +0.0045 World model loss: +0.5977 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5928 Episode_Reward/track_angular_velocity: +49.7090 Episode_Reward/upright: +25.2081 Episode_Reward/leg_posture: -2.4892 Episode_Reward/standstill: -6.1642 Episode_Reward/action_rate_l2: -0.5432 Episode_Reward/joint_torques_l2: -0.1412 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0077 Episode_Reward/unnecessary_wheel_lift: -0.0684 Episode_Reward/rolling_leg_motion: -0.5615 Episode_Reward/obstacle_progress: +3.5353 Episode_Reward/wheel_body_intrusion: -0.2568 Episode_Reward/pair_body_intrusion: -0.3030 Curriculum/terrain_level_mean: +3.9919 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4247 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.5548 Episode_Termination/illegal_nonwheel_contact: 0.0137 ============================================================================== ============================================================================== Learning iteration 33450/50000 Total steps: 137,011,200 Steps per second: 13,215 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1375.3 Actor loss: -9.5104 Critic loss: +0.0051 World model loss: +0.6508 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7485 Episode_Reward/track_angular_velocity: +52.0001 Episode_Reward/upright: +26.3722 Episode_Reward/leg_posture: -2.6327 Episode_Reward/standstill: -6.4425 Episode_Reward/action_rate_l2: -0.5712 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0269 Episode_Reward/unnecessary_wheel_lift: -0.0857 Episode_Reward/rolling_leg_motion: -0.5996 Episode_Reward/obstacle_progress: +3.7023 Episode_Reward/wheel_body_intrusion: -0.2820 Episode_Reward/pair_body_intrusion: -0.3008 Curriculum/terrain_level_mean: +3.9932 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4966 Episode_Termination/fell_over: 0.0483 Episode_Termination/course_exit: 0.4345 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== ============================================================================== Learning iteration 33500/50000 Total steps: 137,216,000 Steps per second: 13,218 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1435.8 Actor loss: -9.4863 Critic loss: +0.0091 World model loss: +0.6500 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.6519 Episode_Reward/track_angular_velocity: +54.3329 Episode_Reward/upright: +27.5243 Episode_Reward/leg_posture: -2.7486 Episode_Reward/standstill: -6.5881 Episode_Reward/action_rate_l2: -0.5995 Episode_Reward/joint_torques_l2: -0.1565 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0196 Episode_Reward/unnecessary_wheel_lift: -0.0921 Episode_Reward/rolling_leg_motion: -0.6300 Episode_Reward/obstacle_progress: +3.7865 Episode_Reward/wheel_body_intrusion: -0.2897 Episode_Reward/pair_body_intrusion: -0.3387 Curriculum/terrain_level_mean: +3.9937 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.5325 Episode_Termination/fell_over: 0.0296 Episode_Termination/course_exit: 0.4260 Episode_Termination/illegal_nonwheel_contact: 0.0118 ============================================================================== ============================================================================== Learning iteration 33550/50000 Total steps: 137,420,800 Steps per second: 13,221 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0877 Mean episode length: 1368.3 Actor loss: -9.4914 Critic loss: +0.0056 World model loss: +0.6079 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4222 Episode_Reward/track_angular_velocity: +51.7776 Episode_Reward/upright: +26.2418 Episode_Reward/leg_posture: -2.5833 Episode_Reward/standstill: -5.5608 Episode_Reward/action_rate_l2: -0.5792 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.0885 Episode_Reward/rolling_leg_motion: -0.5929 Episode_Reward/obstacle_progress: +3.7281 Episode_Reward/wheel_body_intrusion: -0.2622 Episode_Reward/pair_body_intrusion: -0.2941 Curriculum/terrain_level_mean: +3.9941 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4765 Episode_Termination/fell_over: 0.0470 Episode_Termination/course_exit: 0.4564 Episode_Termination/illegal_nonwheel_contact: 0.0470 ============================================================================== ============================================================================== Learning iteration 33600/50000 Total steps: 137,625,600 Steps per second: 13,224 Iteration time: 0.310s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1376.9 Actor loss: -9.5129 Critic loss: +0.0112 World model loss: +0.5821 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6627 Episode_Reward/track_angular_velocity: +52.0496 Episode_Reward/upright: +26.3874 Episode_Reward/leg_posture: -2.5655 Episode_Reward/standstill: -5.5901 Episode_Reward/action_rate_l2: -0.5888 Episode_Reward/joint_torques_l2: -0.1486 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0739 Episode_Reward/rolling_leg_motion: -0.5735 Episode_Reward/obstacle_progress: +3.8455 Episode_Reward/wheel_body_intrusion: -0.2735 Episode_Reward/pair_body_intrusion: -0.3029 Curriculum/terrain_level_mean: +3.9944 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5120 Episode_Termination/fell_over: 0.0240 Episode_Termination/course_exit: 0.4320 Episode_Termination/illegal_nonwheel_contact: 0.0320 ============================================================================== ============================================================================== Learning iteration 33650/50000 Total steps: 137,830,400 Steps per second: 13,227 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1316.0 Actor loss: -9.5013 Critic loss: +0.0051 World model loss: +0.6300 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4302 Episode_Reward/track_angular_velocity: +49.7475 Episode_Reward/upright: +25.1775 Episode_Reward/leg_posture: -2.4778 Episode_Reward/standstill: -5.5937 Episode_Reward/action_rate_l2: -0.5596 Episode_Reward/joint_torques_l2: -0.1419 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0283 Episode_Reward/unnecessary_wheel_lift: -0.0904 Episode_Reward/rolling_leg_motion: -0.5628 Episode_Reward/obstacle_progress: +3.6364 Episode_Reward/wheel_body_intrusion: -0.2642 Episode_Reward/pair_body_intrusion: -0.3172 Curriculum/terrain_level_mean: +3.9951 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0950 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4645 Episode_Termination/fell_over: 0.0581 Episode_Termination/course_exit: 0.4645 Episode_Termination/illegal_nonwheel_contact: 0.0387 ============================================================================== ============================================================================== Learning iteration 33700/50000 Total steps: 138,035,200 Steps per second: 13,230 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1375.0 Actor loss: -9.4901 Critic loss: +0.0074 World model loss: +0.5961 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6929 Episode_Reward/track_angular_velocity: +51.9716 Episode_Reward/upright: +26.3366 Episode_Reward/leg_posture: -2.5953 Episode_Reward/standstill: -5.6379 Episode_Reward/action_rate_l2: -0.5887 Episode_Reward/joint_torques_l2: -0.1499 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0874 Episode_Reward/rolling_leg_motion: -0.5982 Episode_Reward/obstacle_progress: +3.7897 Episode_Reward/wheel_body_intrusion: -0.2683 Episode_Reward/pair_body_intrusion: -0.3110 Curriculum/terrain_level_mean: +3.9958 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.5223 Episode_Termination/fell_over: 0.0318 Episode_Termination/course_exit: 0.4331 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== ============================================================================== Learning iteration 33750/50000 Total steps: 138,240,000 Steps per second: 13,233 Iteration time: 0.310s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0879 Mean episode length: 1354.1 Actor loss: -9.5295 Critic loss: +0.0043 World model loss: +0.5928 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0681 Episode_Reward/track_angular_velocity: +51.2796 Episode_Reward/upright: +25.9560 Episode_Reward/leg_posture: -2.5454 Episode_Reward/standstill: -5.3682 Episode_Reward/action_rate_l2: -0.5719 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0264 Episode_Reward/unnecessary_wheel_lift: -0.0906 Episode_Reward/rolling_leg_motion: -0.5922 Episode_Reward/obstacle_progress: +3.6495 Episode_Reward/wheel_body_intrusion: -0.2573 Episode_Reward/pair_body_intrusion: -0.2974 Curriculum/terrain_level_mean: +3.9961 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.4694 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4626 Episode_Termination/illegal_nonwheel_contact: 0.0612 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[468] ============================================================================== Learning iteration 33800/50000 Total steps: 138,444,800 Steps per second: 13,236 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0879 Mean episode length: 1339.5 Actor loss: -9.4788 Critic loss: +0.0068 World model loss: +0.6192 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7031 Episode_Reward/track_angular_velocity: +50.6779 Episode_Reward/upright: +25.6991 Episode_Reward/leg_posture: -2.4907 Episode_Reward/standstill: -5.6036 Episode_Reward/action_rate_l2: -0.5724 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0052 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5844 Episode_Reward/obstacle_progress: +3.6181 Episode_Reward/wheel_body_intrusion: -0.2486 Episode_Reward/pair_body_intrusion: -0.2852 Curriculum/terrain_level_mean: +3.9993 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.4521 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.5342 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== ============================================================================== Learning iteration 33850/50000 Total steps: 138,649,600 Steps per second: 13,239 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1391.9 Actor loss: -9.4738 Critic loss: +0.0066 World model loss: +0.5936 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4298 Episode_Reward/track_angular_velocity: +52.5707 Episode_Reward/upright: +26.7128 Episode_Reward/leg_posture: -2.5938 Episode_Reward/standstill: -5.9735 Episode_Reward/action_rate_l2: -0.5944 Episode_Reward/joint_torques_l2: -0.1489 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.0781 Episode_Reward/rolling_leg_motion: -0.6097 Episode_Reward/obstacle_progress: +3.6788 Episode_Reward/wheel_body_intrusion: -0.2708 Episode_Reward/pair_body_intrusion: -0.3101 Curriculum/terrain_level_mean: +4.0000 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5355 Episode_Termination/fell_over: 0.0387 Episode_Termination/course_exit: 0.4129 Episode_Termination/illegal_nonwheel_contact: 0.0194 ============================================================================== ============================================================================== Learning iteration 33900/50000 Total steps: 138,854,400 Steps per second: 13,242 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1387.3 Actor loss: -9.5246 Critic loss: +0.0046 World model loss: +0.6267 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4414 Episode_Reward/track_angular_velocity: +52.4849 Episode_Reward/upright: +26.5824 Episode_Reward/leg_posture: -2.5803 Episode_Reward/standstill: -5.9285 Episode_Reward/action_rate_l2: -0.5685 Episode_Reward/joint_torques_l2: -0.1486 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.0896 Episode_Reward/rolling_leg_motion: -0.5994 Episode_Reward/obstacle_progress: +3.7162 Episode_Reward/wheel_body_intrusion: -0.2553 Episode_Reward/pair_body_intrusion: -0.2992 Curriculum/terrain_level_mean: +4.0017 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.5306 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4218 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1671] ============================================================================== Learning iteration 33950/50000 Total steps: 139,059,200 Steps per second: 13,245 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1400.1 Actor loss: -9.5191 Critic loss: +0.0046 World model loss: +0.5795 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8431 Episode_Reward/track_angular_velocity: +52.9455 Episode_Reward/upright: +26.8387 Episode_Reward/leg_posture: -2.6451 Episode_Reward/standstill: -6.1741 Episode_Reward/action_rate_l2: -0.5873 Episode_Reward/joint_torques_l2: -0.1511 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0848 Episode_Reward/rolling_leg_motion: -0.6034 Episode_Reward/obstacle_progress: +3.7657 Episode_Reward/wheel_body_intrusion: -0.2697 Episode_Reward/pair_body_intrusion: -0.3010 Curriculum/terrain_level_mean: +4.0039 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5220 Episode_Termination/fell_over: 0.0126 Episode_Termination/course_exit: 0.4528 Episode_Termination/illegal_nonwheel_contact: 0.0126 ============================================================================== ============================================================================== Learning iteration 34000/50000 Total steps: 139,264,000 Steps per second: 13,248 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1353.0 Actor loss: -9.4818 Critic loss: +0.0099 World model loss: +0.6079 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2214 Episode_Reward/track_angular_velocity: +51.1721 Episode_Reward/upright: +25.9492 Episode_Reward/leg_posture: -2.5717 Episode_Reward/standstill: -5.9645 Episode_Reward/action_rate_l2: -0.5671 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0820 Episode_Reward/rolling_leg_motion: -0.5890 Episode_Reward/obstacle_progress: +3.6617 Episode_Reward/wheel_body_intrusion: -0.2628 Episode_Reward/pair_body_intrusion: -0.2828 Curriculum/terrain_level_mean: +4.0049 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4353 Episode_Termination/fell_over: 0.0176 Episode_Termination/course_exit: 0.5235 Episode_Termination/illegal_nonwheel_contact: 0.0294 ============================================================================== ============================================================================== Learning iteration 34050/50000 Total steps: 139,468,800 Steps per second: 13,251 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1330.1 Actor loss: -9.5157 Critic loss: +0.0047 World model loss: +0.5669 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3945 Episode_Reward/track_angular_velocity: +50.3264 Episode_Reward/upright: +25.4988 Episode_Reward/leg_posture: -2.5173 Episode_Reward/standstill: -5.8730 Episode_Reward/action_rate_l2: -0.5585 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.0757 Episode_Reward/rolling_leg_motion: -0.5795 Episode_Reward/obstacle_progress: +3.5233 Episode_Reward/wheel_body_intrusion: -0.2535 Episode_Reward/pair_body_intrusion: -0.2699 Curriculum/terrain_level_mean: +4.0054 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0907 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4392 Episode_Termination/fell_over: 0.0473 Episode_Termination/course_exit: 0.4932 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== ============================================================================== Learning iteration 34100/50000 Total steps: 139,673,600 Steps per second: 13,254 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1375.4 Actor loss: -9.5303 Critic loss: +0.0030 World model loss: +0.6208 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9433 Episode_Reward/track_angular_velocity: +52.0465 Episode_Reward/upright: +26.3625 Episode_Reward/leg_posture: -2.5465 Episode_Reward/standstill: -5.9915 Episode_Reward/action_rate_l2: -0.5744 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0215 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5976 Episode_Reward/obstacle_progress: +3.7161 Episode_Reward/wheel_body_intrusion: -0.2599 Episode_Reward/pair_body_intrusion: -0.2860 Curriculum/terrain_level_mean: +4.0068 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.5091 Episode_Termination/fell_over: 0.0242 Episode_Termination/course_exit: 0.4424 Episode_Termination/illegal_nonwheel_contact: 0.0364 ============================================================================== ============================================================================== Learning iteration 34150/50000 Total steps: 139,878,400 Steps per second: 13,257 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1352.7 Actor loss: -9.5081 Critic loss: +0.0042 World model loss: +0.5854 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1592 Episode_Reward/track_angular_velocity: +51.2067 Episode_Reward/upright: +25.9441 Episode_Reward/leg_posture: -2.5065 Episode_Reward/standstill: -5.9092 Episode_Reward/action_rate_l2: -0.5571 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0730 Episode_Reward/rolling_leg_motion: -0.5780 Episode_Reward/obstacle_progress: +3.7246 Episode_Reward/wheel_body_intrusion: -0.2655 Episode_Reward/pair_body_intrusion: -0.2700 Curriculum/terrain_level_mean: +4.0093 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4865 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.4865 Episode_Termination/illegal_nonwheel_contact: 0.0203 ============================================================================== ============================================================================== Learning iteration 34200/50000 Total steps: 140,083,200 Steps per second: 13,260 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1360.2 Actor loss: -9.5070 Critic loss: +0.0052 World model loss: +0.6155 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3500 Episode_Reward/track_angular_velocity: +51.4660 Episode_Reward/upright: +26.0371 Episode_Reward/leg_posture: -2.5992 Episode_Reward/standstill: -5.9595 Episode_Reward/action_rate_l2: -0.5640 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0133 Episode_Reward/unnecessary_wheel_lift: -0.0888 Episode_Reward/rolling_leg_motion: -0.6013 Episode_Reward/obstacle_progress: +3.6926 Episode_Reward/wheel_body_intrusion: -0.2768 Episode_Reward/pair_body_intrusion: -0.2886 Curriculum/terrain_level_mean: +4.0110 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4819 Episode_Termination/fell_over: 0.0241 Episode_Termination/course_exit: 0.4880 Episode_Termination/illegal_nonwheel_contact: 0.0241 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[675] ============================================================================== Learning iteration 34250/50000 Total steps: 140,288,000 Steps per second: 13,263 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1347.8 Actor loss: -9.5026 Critic loss: +0.0049 World model loss: +0.5827 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0553 Episode_Reward/track_angular_velocity: +51.0135 Episode_Reward/upright: +25.8020 Episode_Reward/leg_posture: -2.5519 Episode_Reward/standstill: -5.8135 Episode_Reward/action_rate_l2: -0.5650 Episode_Reward/joint_torques_l2: -0.1431 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0105 Episode_Reward/unnecessary_wheel_lift: -0.0806 Episode_Reward/rolling_leg_motion: -0.5957 Episode_Reward/obstacle_progress: +3.6389 Episode_Reward/wheel_body_intrusion: -0.2610 Episode_Reward/pair_body_intrusion: -0.2663 Curriculum/terrain_level_mean: +4.0159 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.4737 Episode_Termination/fell_over: 0.0066 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== ============================================================================== Learning iteration 34300/50000 Total steps: 140,492,800 Steps per second: 13,266 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1312.2 Actor loss: -9.5155 Critic loss: +0.0047 World model loss: +0.5656 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6732 Episode_Reward/track_angular_velocity: +49.7002 Episode_Reward/upright: +25.1274 Episode_Reward/leg_posture: -2.4420 Episode_Reward/standstill: -5.7447 Episode_Reward/action_rate_l2: -0.5439 Episode_Reward/joint_torques_l2: -0.1401 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0205 Episode_Reward/unnecessary_wheel_lift: -0.0809 Episode_Reward/rolling_leg_motion: -0.5716 Episode_Reward/obstacle_progress: +3.5191 Episode_Reward/wheel_body_intrusion: -0.2437 Episode_Reward/pair_body_intrusion: -0.2712 Curriculum/terrain_level_mean: +4.0195 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1321 Episode_Termination/time_out: 0.4658 Episode_Termination/fell_over: 0.0248 Episode_Termination/course_exit: 0.4845 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== ============================================================================== Learning iteration 34350/50000 Total steps: 140,697,600 Steps per second: 13,269 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1372.3 Actor loss: -9.5060 Critic loss: +0.0047 World model loss: +0.5913 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6934 Episode_Reward/track_angular_velocity: +51.8948 Episode_Reward/upright: +26.2554 Episode_Reward/leg_posture: -2.5499 Episode_Reward/standstill: -6.0282 Episode_Reward/action_rate_l2: -0.5715 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0797 Episode_Reward/rolling_leg_motion: -0.5890 Episode_Reward/obstacle_progress: +3.6645 Episode_Reward/wheel_body_intrusion: -0.2578 Episode_Reward/pair_body_intrusion: -0.2776 Curriculum/terrain_level_mean: +4.0208 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5827 Episode_Termination/fell_over: 0.0144 Episode_Termination/course_exit: 0.3885 Episode_Termination/illegal_nonwheel_contact: 0.0288 ============================================================================== ============================================================================== Learning iteration 34400/50000 Total steps: 140,902,400 Steps per second: 13,272 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1363.2 Actor loss: -9.5238 Critic loss: +0.0054 World model loss: +0.5857 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3554 Episode_Reward/track_angular_velocity: +51.5671 Episode_Reward/upright: +26.0999 Episode_Reward/leg_posture: -2.5708 Episode_Reward/standstill: -5.6773 Episode_Reward/action_rate_l2: -0.5676 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0240 Episode_Reward/unnecessary_wheel_lift: -0.0735 Episode_Reward/rolling_leg_motion: -0.5847 Episode_Reward/obstacle_progress: +3.7139 Episode_Reward/wheel_body_intrusion: -0.2541 Episode_Reward/pair_body_intrusion: -0.2848 Curriculum/terrain_level_mean: +4.0242 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4485 Episode_Termination/fell_over: 0.0061 Episode_Termination/course_exit: 0.4970 Episode_Termination/illegal_nonwheel_contact: 0.0485 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1164] ============================================================================== Learning iteration 34450/50000 Total steps: 141,107,200 Steps per second: 13,275 Iteration time: 0.309s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1366.7 Actor loss: -9.5202 Critic loss: +0.0053 World model loss: +0.5601 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5433 Episode_Reward/track_angular_velocity: +51.7009 Episode_Reward/upright: +26.1701 Episode_Reward/leg_posture: -2.5605 Episode_Reward/standstill: -5.6558 Episode_Reward/action_rate_l2: -0.5748 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0772 Episode_Reward/rolling_leg_motion: -0.5864 Episode_Reward/obstacle_progress: +3.7732 Episode_Reward/wheel_body_intrusion: -0.2804 Episode_Reward/pair_body_intrusion: -0.3065 Curriculum/terrain_level_mean: +4.0271 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.4717 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0126 ============================================================================== ============================================================================== Learning iteration 34500/50000 Total steps: 141,312,000 Steps per second: 13,278 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1380.9 Actor loss: -9.5222 Critic loss: +0.0031 World model loss: +0.5995 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9121 Episode_Reward/track_angular_velocity: +52.1986 Episode_Reward/upright: +26.4599 Episode_Reward/leg_posture: -2.6004 Episode_Reward/standstill: -5.9149 Episode_Reward/action_rate_l2: -0.5722 Episode_Reward/joint_torques_l2: -0.1486 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.0917 Episode_Reward/rolling_leg_motion: -0.6184 Episode_Reward/obstacle_progress: +3.8658 Episode_Reward/wheel_body_intrusion: -0.2738 Episode_Reward/pair_body_intrusion: -0.2879 Curriculum/terrain_level_mean: +4.0283 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4825 Episode_Termination/fell_over: 0.0280 Episode_Termination/course_exit: 0.4825 Episode_Termination/illegal_nonwheel_contact: 0.0210 ============================================================================== ============================================================================== Learning iteration 34550/50000 Total steps: 141,516,800 Steps per second: 13,281 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0880 Mean episode length: 1389.9 Actor loss: -9.5296 Critic loss: +0.0041 World model loss: +0.5875 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3226 Episode_Reward/track_angular_velocity: +52.5693 Episode_Reward/upright: +26.6376 Episode_Reward/leg_posture: -2.5953 Episode_Reward/standstill: -5.5843 Episode_Reward/action_rate_l2: -0.5760 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0190 Episode_Reward/unnecessary_wheel_lift: -0.0870 Episode_Reward/rolling_leg_motion: -0.6011 Episode_Reward/obstacle_progress: +3.9657 Episode_Reward/wheel_body_intrusion: -0.2597 Episode_Reward/pair_body_intrusion: -0.2839 Curriculum/terrain_level_mean: +4.0305 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5097 Episode_Termination/fell_over: 0.0323 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0258 ============================================================================== ============================================================================== Learning iteration 34600/50000 Total steps: 141,721,600 Steps per second: 13,284 Iteration time: 0.308s Collection time: 0.075s Learning time: 0.186s Mean reward: +0.0880 Mean episode length: 1417.1 Actor loss: -9.5365 Critic loss: +0.0035 World model loss: +0.5819 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2229 Episode_Reward/track_angular_velocity: +53.5767 Episode_Reward/upright: +27.1504 Episode_Reward/leg_posture: -2.6176 Episode_Reward/standstill: -5.7443 Episode_Reward/action_rate_l2: -0.5887 Episode_Reward/joint_torques_l2: -0.1521 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0111 Episode_Reward/unnecessary_wheel_lift: -0.0781 Episode_Reward/rolling_leg_motion: -0.6066 Episode_Reward/obstacle_progress: +4.0668 Episode_Reward/wheel_body_intrusion: -0.2714 Episode_Reward/pair_body_intrusion: -0.2966 Curriculum/terrain_level_mean: +4.0327 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.5255 Episode_Termination/fell_over: 0.0073 Episode_Termination/course_exit: 0.4453 Episode_Termination/illegal_nonwheel_contact: 0.0292 ============================================================================== ============================================================================== Learning iteration 34650/50000 Total steps: 141,926,400 Steps per second: 13,287 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1312.9 Actor loss: -9.5107 Critic loss: +0.0050 World model loss: +0.5562 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4921 Episode_Reward/track_angular_velocity: +49.6014 Episode_Reward/upright: +25.1531 Episode_Reward/leg_posture: -2.4559 Episode_Reward/standstill: -5.6934 Episode_Reward/action_rate_l2: -0.5409 Episode_Reward/joint_torques_l2: -0.1415 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0103 Episode_Reward/unnecessary_wheel_lift: -0.0769 Episode_Reward/rolling_leg_motion: -0.5611 Episode_Reward/obstacle_progress: +3.7468 Episode_Reward/wheel_body_intrusion: -0.2464 Episode_Reward/pair_body_intrusion: -0.2575 Curriculum/terrain_level_mean: +4.0352 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4662 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.5038 Episode_Termination/illegal_nonwheel_contact: 0.0150 ============================================================================== ============================================================================== Learning iteration 34700/50000 Total steps: 142,131,200 Steps per second: 13,290 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0880 Mean episode length: 1355.0 Actor loss: -9.5591 Critic loss: +0.0037 World model loss: +0.6102 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1049 Episode_Reward/track_angular_velocity: +51.2229 Episode_Reward/upright: +25.9955 Episode_Reward/leg_posture: -2.5251 Episode_Reward/standstill: -5.5622 Episode_Reward/action_rate_l2: -0.5511 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0864 Episode_Reward/rolling_leg_motion: -0.5813 Episode_Reward/obstacle_progress: +3.8861 Episode_Reward/wheel_body_intrusion: -0.2438 Episode_Reward/pair_body_intrusion: -0.2796 Curriculum/terrain_level_mean: +4.0388 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4812 Episode_Termination/fell_over: 0.0301 Episode_Termination/course_exit: 0.4887 Episode_Termination/illegal_nonwheel_contact: 0.0150 ============================================================================== ============================================================================== Learning iteration 34750/50000 Total steps: 142,336,000 Steps per second: 13,293 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0881 Mean episode length: 1400.8 Actor loss: -9.5228 Critic loss: +0.0042 World model loss: +0.6045 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7953 Episode_Reward/track_angular_velocity: +52.9604 Episode_Reward/upright: +26.9088 Episode_Reward/leg_posture: -2.6043 Episode_Reward/standstill: -5.5532 Episode_Reward/action_rate_l2: -0.5719 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0055 Episode_Reward/unnecessary_wheel_lift: -0.0817 Episode_Reward/rolling_leg_motion: -0.6022 Episode_Reward/obstacle_progress: +3.9439 Episode_Reward/wheel_body_intrusion: -0.2548 Episode_Reward/pair_body_intrusion: -0.3072 Curriculum/terrain_level_mean: +4.0417 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0067 Episode_Termination/course_exit: 0.4832 Episode_Termination/illegal_nonwheel_contact: 0.0067 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[553] ============================================================================== Learning iteration 34800/50000 Total steps: 142,540,800 Steps per second: 13,296 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0881 Mean episode length: 1370.5 Actor loss: -9.5167 Critic loss: +0.0072 World model loss: +0.6119 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7187 Episode_Reward/track_angular_velocity: +51.8622 Episode_Reward/upright: +26.3290 Episode_Reward/leg_posture: -2.5283 Episode_Reward/standstill: -5.5799 Episode_Reward/action_rate_l2: -0.5661 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0852 Episode_Reward/rolling_leg_motion: -0.6029 Episode_Reward/obstacle_progress: +3.7997 Episode_Reward/wheel_body_intrusion: -0.2458 Episode_Reward/pair_body_intrusion: -0.2687 Curriculum/terrain_level_mean: +4.0444 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.5062 Episode_Termination/fell_over: 0.0250 Episode_Termination/course_exit: 0.4625 Episode_Termination/illegal_nonwheel_contact: 0.0187 ============================================================================== ============================================================================== Learning iteration 34850/50000 Total steps: 142,745,600 Steps per second: 13,298 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0883 Mean episode length: 1301.1 Actor loss: -9.5376 Critic loss: +0.0040 World model loss: +0.6154 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4207 Episode_Reward/track_angular_velocity: +49.2626 Episode_Reward/upright: +25.0253 Episode_Reward/leg_posture: -2.3859 Episode_Reward/standstill: -5.2105 Episode_Reward/action_rate_l2: -0.5348 Episode_Reward/joint_torques_l2: -0.1380 Episode_Reward/dof_pos_limits: -0.0001 Episode_Reward/terminated: -0.0127 Episode_Reward/unnecessary_wheel_lift: -0.0693 Episode_Reward/rolling_leg_motion: -0.5608 Episode_Reward/obstacle_progress: +3.6185 Episode_Reward/wheel_body_intrusion: -0.2316 Episode_Reward/pair_body_intrusion: -0.2700 Curriculum/terrain_level_mean: +4.0454 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1323 Episode_Termination/time_out: 0.4567 Episode_Termination/fell_over: 0.0079 Episode_Termination/course_exit: 0.5039 Episode_Termination/illegal_nonwheel_contact: 0.0315 ============================================================================== ============================================================================== Learning iteration 34900/50000 Total steps: 142,950,400 Steps per second: 13,301 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1308.6 Actor loss: -9.5084 Critic loss: +0.0035 World model loss: +0.5561 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4954 Episode_Reward/track_angular_velocity: +49.4959 Episode_Reward/upright: +25.1308 Episode_Reward/leg_posture: -2.3840 Episode_Reward/standstill: -5.4757 Episode_Reward/action_rate_l2: -0.5523 Episode_Reward/joint_torques_l2: -0.1396 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0307 Episode_Reward/unnecessary_wheel_lift: -0.0860 Episode_Reward/rolling_leg_motion: -0.5690 Episode_Reward/obstacle_progress: +3.6836 Episode_Reward/wheel_body_intrusion: -0.2325 Episode_Reward/pair_body_intrusion: -0.2641 Curriculum/terrain_level_mean: +4.0461 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.4506 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4753 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2559] ============================================================================== Learning iteration 34950/50000 Total steps: 143,155,200 Steps per second: 13,304 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1314.6 Actor loss: -9.5328 Critic loss: +0.0034 World model loss: +0.6309 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7021 Episode_Reward/track_angular_velocity: +49.7155 Episode_Reward/upright: +25.2477 Episode_Reward/leg_posture: -2.4192 Episode_Reward/standstill: -5.6061 Episode_Reward/action_rate_l2: -0.5500 Episode_Reward/joint_torques_l2: -0.1397 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0282 Episode_Reward/unnecessary_wheel_lift: -0.0944 Episode_Reward/rolling_leg_motion: -0.5659 Episode_Reward/obstacle_progress: +3.7532 Episode_Reward/wheel_body_intrusion: -0.2355 Episode_Reward/pair_body_intrusion: -0.2831 Curriculum/terrain_level_mean: +4.0466 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4172 Episode_Termination/fell_over: 0.0307 Episode_Termination/course_exit: 0.5153 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 35000/50000 Total steps: 143,360,000 Steps per second: 13,307 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0882 Mean episode length: 1364.4 Actor loss: -9.5425 Critic loss: +0.0048 World model loss: +0.6068 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5488 Episode_Reward/track_angular_velocity: +51.5951 Episode_Reward/upright: +26.2323 Episode_Reward/leg_posture: -2.5270 Episode_Reward/standstill: -5.4660 Episode_Reward/action_rate_l2: -0.5677 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: +0.0000 Episode_Reward/unnecessary_wheel_lift: -0.0860 Episode_Reward/rolling_leg_motion: -0.5913 Episode_Reward/obstacle_progress: +3.8668 Episode_Reward/wheel_body_intrusion: -0.2559 Episode_Reward/pair_body_intrusion: -0.2967 Curriculum/terrain_level_mean: +4.0498 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4641 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5359 Episode_Termination/illegal_nonwheel_contact: 0.0000 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[480] ============================================================================== Learning iteration 35050/50000 Total steps: 143,564,800 Steps per second: 13,310 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1278.8 Actor loss: -9.5481 Critic loss: +0.0051 World model loss: +0.5548 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.3828 Episode_Reward/track_angular_velocity: +48.3301 Episode_Reward/upright: +24.5550 Episode_Reward/leg_posture: -2.3921 Episode_Reward/standstill: -5.5661 Episode_Reward/action_rate_l2: -0.5257 Episode_Reward/joint_torques_l2: -0.1380 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0225 Episode_Reward/unnecessary_wheel_lift: -0.0934 Episode_Reward/rolling_leg_motion: -0.5520 Episode_Reward/obstacle_progress: +3.5933 Episode_Reward/wheel_body_intrusion: -0.2577 Episode_Reward/pair_body_intrusion: -0.2979 Curriculum/terrain_level_mean: +4.0503 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.3988 Episode_Termination/fell_over: 0.0231 Episode_Termination/course_exit: 0.5491 Episode_Termination/illegal_nonwheel_contact: 0.0405 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1143] ============================================================================== Learning iteration 35100/50000 Total steps: 143,769,600 Steps per second: 13,312 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1325.9 Actor loss: -9.5479 Critic loss: +0.0061 World model loss: +0.5950 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0509 Episode_Reward/track_angular_velocity: +50.1066 Episode_Reward/upright: +25.4155 Episode_Reward/leg_posture: -2.4958 Episode_Reward/standstill: -5.9700 Episode_Reward/action_rate_l2: -0.5493 Episode_Reward/joint_torques_l2: -0.1435 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0285 Episode_Reward/unnecessary_wheel_lift: -0.0915 Episode_Reward/rolling_leg_motion: -0.5776 Episode_Reward/obstacle_progress: +3.7519 Episode_Reward/wheel_body_intrusion: -0.2581 Episode_Reward/pair_body_intrusion: -0.3181 Curriculum/terrain_level_mean: +4.0503 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4333 Episode_Termination/fell_over: 0.0400 Episode_Termination/course_exit: 0.4933 Episode_Termination/illegal_nonwheel_contact: 0.0400 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_035100.mp4 RECOVERY reset 1 invalid environments; first IDs=[584] ============================================================================== Learning iteration 35150/50000 Total steps: 143,974,400 Steps per second: 13,308 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1410.0 Actor loss: -9.5621 Critic loss: +0.0049 World model loss: +0.6119 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0089 Episode_Reward/track_angular_velocity: +53.2935 Episode_Reward/upright: +27.0541 Episode_Reward/leg_posture: -2.6297 Episode_Reward/standstill: -6.0779 Episode_Reward/action_rate_l2: -0.5787 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0110 Episode_Reward/unnecessary_wheel_lift: -0.0891 Episode_Reward/rolling_leg_motion: -0.5866 Episode_Reward/obstacle_progress: +4.0194 Episode_Reward/wheel_body_intrusion: -0.2665 Episode_Reward/pair_body_intrusion: -0.2911 Curriculum/terrain_level_mean: +4.0515 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0227 Episode_Termination/course_exit: 0.4697 Episode_Termination/illegal_nonwheel_contact: 0.0227 ============================================================================== ============================================================================== Learning iteration 35200/50000 Total steps: 144,179,200 Steps per second: 13,311 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1354.4 Actor loss: -9.4804 Critic loss: +0.0054 World model loss: +0.6003 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9897 Episode_Reward/track_angular_velocity: +51.1906 Episode_Reward/upright: +25.9735 Episode_Reward/leg_posture: -2.5497 Episode_Reward/standstill: -5.9074 Episode_Reward/action_rate_l2: -0.5578 Episode_Reward/joint_torques_l2: -0.1451 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.1062 Episode_Reward/rolling_leg_motion: -0.5922 Episode_Reward/obstacle_progress: +3.8107 Episode_Reward/wheel_body_intrusion: -0.2501 Episode_Reward/pair_body_intrusion: -0.2600 Curriculum/terrain_level_mean: +4.0537 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4438 Episode_Termination/fell_over: 0.0296 Episode_Termination/course_exit: 0.5207 Episode_Termination/illegal_nonwheel_contact: 0.0296 ============================================================================== ============================================================================== Learning iteration 35250/50000 Total steps: 144,384,000 Steps per second: 13,314 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1353.2 Actor loss: -9.5315 Critic loss: +0.0035 World model loss: +0.5887 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9889 Episode_Reward/track_angular_velocity: +51.1690 Episode_Reward/upright: +25.9559 Episode_Reward/leg_posture: -2.5465 Episode_Reward/standstill: -6.0101 Episode_Reward/action_rate_l2: -0.5582 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.1019 Episode_Reward/rolling_leg_motion: -0.5982 Episode_Reward/obstacle_progress: +3.8272 Episode_Reward/wheel_body_intrusion: -0.2504 Episode_Reward/pair_body_intrusion: -0.3077 Curriculum/terrain_level_mean: +4.0537 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.4912 Episode_Termination/fell_over: 0.0175 Episode_Termination/course_exit: 0.4678 Episode_Termination/illegal_nonwheel_contact: 0.0292 ============================================================================== ============================================================================== Learning iteration 35300/50000 Total steps: 144,588,800 Steps per second: 13,317 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0880 Mean episode length: 1365.9 Actor loss: -9.5439 Critic loss: +0.0051 World model loss: +0.6017 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5649 Episode_Reward/track_angular_velocity: +51.6946 Episode_Reward/upright: +26.2097 Episode_Reward/leg_posture: -2.5549 Episode_Reward/standstill: -5.6176 Episode_Reward/action_rate_l2: -0.5693 Episode_Reward/joint_torques_l2: -0.1476 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.0921 Episode_Reward/rolling_leg_motion: -0.5908 Episode_Reward/obstacle_progress: +3.9215 Episode_Reward/wheel_body_intrusion: -0.2461 Episode_Reward/pair_body_intrusion: -0.3079 Curriculum/terrain_level_mean: +4.0552 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4832 Episode_Termination/fell_over: 0.0268 Episode_Termination/course_exit: 0.4631 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 35350/50000 Total steps: 144,793,600 Steps per second: 13,320 Iteration time: 0.308s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0882 Mean episode length: 1294.2 Actor loss: -9.5263 Critic loss: +0.0059 World model loss: +0.5861 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1163 Episode_Reward/track_angular_velocity: +48.9519 Episode_Reward/upright: +24.8519 Episode_Reward/leg_posture: -2.4083 Episode_Reward/standstill: -5.3543 Episode_Reward/action_rate_l2: -0.5315 Episode_Reward/joint_torques_l2: -0.1388 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0051 Episode_Reward/unnecessary_wheel_lift: -0.0903 Episode_Reward/rolling_leg_motion: -0.5626 Episode_Reward/obstacle_progress: +3.7487 Episode_Reward/wheel_body_intrusion: -0.2339 Episode_Reward/pair_body_intrusion: -0.2335 Curriculum/terrain_level_mean: +4.0591 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.3919 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.5946 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2846] ============================================================================== Learning iteration 35400/50000 Total steps: 144,998,400 Steps per second: 13,322 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0883 Mean episode length: 1366.0 Actor loss: -9.5245 Critic loss: +0.0054 World model loss: +0.5840 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6583 Episode_Reward/track_angular_velocity: +51.6646 Episode_Reward/upright: +26.1856 Episode_Reward/leg_posture: -2.5654 Episode_Reward/standstill: -5.5080 Episode_Reward/action_rate_l2: -0.5645 Episode_Reward/joint_torques_l2: -0.1464 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0762 Episode_Reward/rolling_leg_motion: -0.5855 Episode_Reward/obstacle_progress: +4.0241 Episode_Reward/wheel_body_intrusion: -0.2544 Episode_Reward/pair_body_intrusion: -0.2510 Curriculum/terrain_level_mean: +4.0615 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5139 Episode_Termination/fell_over: 0.0069 Episode_Termination/course_exit: 0.4583 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== ============================================================================== Learning iteration 35450/50000 Total steps: 145,203,200 Steps per second: 13,325 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1377.4 Actor loss: -9.5144 Critic loss: +0.0044 World model loss: +0.5902 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9417 Episode_Reward/track_angular_velocity: +52.0670 Episode_Reward/upright: +26.3906 Episode_Reward/leg_posture: -2.5776 Episode_Reward/standstill: -5.7672 Episode_Reward/action_rate_l2: -0.5643 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0269 Episode_Reward/unnecessary_wheel_lift: -0.0925 Episode_Reward/rolling_leg_motion: -0.6039 Episode_Reward/obstacle_progress: +4.0148 Episode_Reward/wheel_body_intrusion: -0.2756 Episode_Reward/pair_body_intrusion: -0.3114 Curriculum/terrain_level_mean: +4.0625 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4906 Episode_Termination/fell_over: 0.0126 Episode_Termination/course_exit: 0.4465 Episode_Termination/illegal_nonwheel_contact: 0.0629 ============================================================================== ============================================================================== Learning iteration 35500/50000 Total steps: 145,408,000 Steps per second: 13,328 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0880 Mean episode length: 1299.9 Actor loss: -9.5205 Critic loss: +0.0058 World model loss: +0.6090 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2951 Episode_Reward/track_angular_velocity: +49.1894 Episode_Reward/upright: +24.9268 Episode_Reward/leg_posture: -2.4565 Episode_Reward/standstill: -5.4378 Episode_Reward/action_rate_l2: -0.5299 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0228 Episode_Reward/unnecessary_wheel_lift: -0.0807 Episode_Reward/rolling_leg_motion: -0.5602 Episode_Reward/obstacle_progress: +3.6751 Episode_Reward/wheel_body_intrusion: -0.2477 Episode_Reward/pair_body_intrusion: -0.2750 Curriculum/terrain_level_mean: +4.0637 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4145 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.5263 Episode_Termination/illegal_nonwheel_contact: 0.0461 ============================================================================== ============================================================================== Learning iteration 35550/50000 Total steps: 145,612,800 Steps per second: 13,330 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0879 Mean episode length: 1350.6 Actor loss: -9.5293 Critic loss: +0.0071 World model loss: +0.6240 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2393 Episode_Reward/track_angular_velocity: +51.1250 Episode_Reward/upright: +25.8901 Episode_Reward/leg_posture: -2.5661 Episode_Reward/standstill: -5.9189 Episode_Reward/action_rate_l2: -0.5455 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0887 Episode_Reward/rolling_leg_motion: -0.5861 Episode_Reward/obstacle_progress: +3.8439 Episode_Reward/wheel_body_intrusion: -0.2612 Episode_Reward/pair_body_intrusion: -0.2908 Curriculum/terrain_level_mean: +4.0664 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0899 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.5238 Episode_Termination/illegal_nonwheel_contact: 0.0272 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2495] RECOVERY reset 1 invalid environments; first IDs=[2087] ============================================================================== Learning iteration 35600/50000 Total steps: 145,817,600 Steps per second: 13,333 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1345.0 Actor loss: -9.5218 Critic loss: +0.0067 World model loss: +0.5927 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0447 Episode_Reward/track_angular_velocity: +50.9061 Episode_Reward/upright: +25.7979 Episode_Reward/leg_posture: -2.5833 Episode_Reward/standstill: -5.9602 Episode_Reward/action_rate_l2: -0.5507 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.0792 Episode_Reward/rolling_leg_motion: -0.5801 Episode_Reward/obstacle_progress: +3.8197 Episode_Reward/wheel_body_intrusion: -0.2499 Episode_Reward/pair_body_intrusion: -0.3122 Curriculum/terrain_level_mean: +4.0684 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0898 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4467 Episode_Termination/fell_over: 0.0267 Episode_Termination/course_exit: 0.5133 Episode_Termination/illegal_nonwheel_contact: 0.0133 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[495] ============================================================================== Learning iteration 35650/50000 Total steps: 146,022,400 Steps per second: 13,336 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1349.0 Actor loss: -9.5546 Critic loss: +0.0043 World model loss: +0.5709 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0623 Episode_Reward/track_angular_velocity: +51.0516 Episode_Reward/upright: +25.8722 Episode_Reward/leg_posture: -2.5780 Episode_Reward/standstill: -6.1706 Episode_Reward/action_rate_l2: -0.5571 Episode_Reward/joint_torques_l2: -0.1462 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0343 Episode_Reward/unnecessary_wheel_lift: -0.0819 Episode_Reward/rolling_leg_motion: -0.5883 Episode_Reward/obstacle_progress: +3.7598 Episode_Reward/wheel_body_intrusion: -0.2523 Episode_Reward/pair_body_intrusion: -0.3026 Curriculum/terrain_level_mean: +4.0669 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4354 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4762 Episode_Termination/illegal_nonwheel_contact: 0.0748 ============================================================================== ============================================================================== Learning iteration 35700/50000 Total steps: 146,227,200 Steps per second: 13,339 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1373.7 Actor loss: -9.5388 Critic loss: +0.0043 World model loss: +0.5951 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8782 Episode_Reward/track_angular_velocity: +51.9878 Episode_Reward/upright: +26.3539 Episode_Reward/leg_posture: -2.6072 Episode_Reward/standstill: -6.4606 Episode_Reward/action_rate_l2: -0.5743 Episode_Reward/joint_torques_l2: -0.1492 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0888 Episode_Reward/rolling_leg_motion: -0.5969 Episode_Reward/obstacle_progress: +3.8322 Episode_Reward/wheel_body_intrusion: -0.2481 Episode_Reward/pair_body_intrusion: -0.2860 Curriculum/terrain_level_mean: +4.0696 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.5089 Episode_Termination/fell_over: 0.0178 Episode_Termination/course_exit: 0.4497 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== ============================================================================== Learning iteration 35750/50000 Total steps: 146,432,000 Steps per second: 13,342 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1335.7 Actor loss: -9.5356 Critic loss: +0.0043 World model loss: +0.5556 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6584 Episode_Reward/track_angular_velocity: +50.5533 Episode_Reward/upright: +25.6378 Episode_Reward/leg_posture: -2.5481 Episode_Reward/standstill: -5.7775 Episode_Reward/action_rate_l2: -0.5548 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0261 Episode_Reward/unnecessary_wheel_lift: -0.0888 Episode_Reward/rolling_leg_motion: -0.5763 Episode_Reward/obstacle_progress: +3.7283 Episode_Reward/wheel_body_intrusion: -0.2428 Episode_Reward/pair_body_intrusion: -0.2604 Curriculum/terrain_level_mean: +4.0698 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4177 Episode_Termination/fell_over: 0.0380 Episode_Termination/course_exit: 0.5190 Episode_Termination/illegal_nonwheel_contact: 0.0316 ============================================================================== ============================================================================== Learning iteration 35800/50000 Total steps: 146,636,800 Steps per second: 13,344 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1334.1 Actor loss: -9.5233 Critic loss: +0.0050 World model loss: +0.5773 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4831 Episode_Reward/track_angular_velocity: +50.4812 Episode_Reward/upright: +25.6550 Episode_Reward/leg_posture: -2.5054 Episode_Reward/standstill: -5.8309 Episode_Reward/action_rate_l2: -0.5608 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.0779 Episode_Reward/rolling_leg_motion: -0.5759 Episode_Reward/obstacle_progress: +3.7568 Episode_Reward/wheel_body_intrusion: -0.2269 Episode_Reward/pair_body_intrusion: -0.2836 Curriculum/terrain_level_mean: +4.0732 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4538 Episode_Termination/fell_over: 0.0154 Episode_Termination/course_exit: 0.5154 Episode_Termination/illegal_nonwheel_contact: 0.0308 ============================================================================== ============================================================================== Learning iteration 35850/50000 Total steps: 146,841,600 Steps per second: 13,347 Iteration time: 0.307s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1381.9 Actor loss: -9.5324 Critic loss: +0.0049 World model loss: +0.6145 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1494 Episode_Reward/track_angular_velocity: +52.2726 Episode_Reward/upright: +26.5360 Episode_Reward/leg_posture: -2.5877 Episode_Reward/standstill: -6.2813 Episode_Reward/action_rate_l2: -0.5850 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0054 Episode_Reward/unnecessary_wheel_lift: -0.0771 Episode_Reward/rolling_leg_motion: -0.5882 Episode_Reward/obstacle_progress: +3.8785 Episode_Reward/wheel_body_intrusion: -0.2567 Episode_Reward/pair_body_intrusion: -0.2845 Curriculum/terrain_level_mean: +4.0754 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4865 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0135 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1296] ============================================================================== Learning iteration 35900/50000 Total steps: 147,046,400 Steps per second: 13,350 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1384.6 Actor loss: -9.5080 Critic loss: +0.0075 World model loss: +0.5948 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2864 Episode_Reward/track_angular_velocity: +52.4142 Episode_Reward/upright: +26.5769 Episode_Reward/leg_posture: -2.5937 Episode_Reward/standstill: -6.1867 Episode_Reward/action_rate_l2: -0.5906 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0270 Episode_Reward/unnecessary_wheel_lift: -0.0887 Episode_Reward/rolling_leg_motion: -0.6008 Episode_Reward/obstacle_progress: +3.8404 Episode_Reward/wheel_body_intrusion: -0.2465 Episode_Reward/pair_body_intrusion: -0.2780 Curriculum/terrain_level_mean: +4.0754 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1322 Episode_Termination/time_out: 0.4702 Episode_Termination/fell_over: 0.0265 Episode_Termination/course_exit: 0.4636 Episode_Termination/illegal_nonwheel_contact: 0.0662 ============================================================================== ============================================================================== Learning iteration 35950/50000 Total steps: 147,251,200 Steps per second: 13,353 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1349.3 Actor loss: -9.5379 Critic loss: +0.0045 World model loss: +0.5618 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0283 Episode_Reward/track_angular_velocity: +51.0648 Episode_Reward/upright: +25.9191 Episode_Reward/leg_posture: -2.5506 Episode_Reward/standstill: -5.8148 Episode_Reward/action_rate_l2: -0.5596 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0158 Episode_Reward/unnecessary_wheel_lift: -0.0842 Episode_Reward/rolling_leg_motion: -0.5770 Episode_Reward/obstacle_progress: +3.7483 Episode_Reward/wheel_body_intrusion: -0.2577 Episode_Reward/pair_body_intrusion: -0.2768 Curriculum/terrain_level_mean: +4.0781 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1328 Episode_Termination/time_out: 0.4643 Episode_Termination/fell_over: 0.0238 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0298 ============================================================================== ============================================================================== Learning iteration 36000/50000 Total steps: 147,456,000 Steps per second: 13,356 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0879 Mean episode length: 1403.3 Actor loss: -9.5166 Critic loss: +0.0126 World model loss: +0.5730 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1035 Episode_Reward/track_angular_velocity: +53.1383 Episode_Reward/upright: +26.9791 Episode_Reward/leg_posture: -2.6380 Episode_Reward/standstill: -6.0969 Episode_Reward/action_rate_l2: -0.5758 Episode_Reward/joint_torques_l2: -0.1508 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0801 Episode_Reward/rolling_leg_motion: -0.6101 Episode_Reward/obstacle_progress: +3.8679 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.2816 Curriculum/terrain_level_mean: +4.0801 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0901 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.5333 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.4333 Episode_Termination/illegal_nonwheel_contact: 0.0267 ============================================================================== ============================================================================== Learning iteration 36050/50000 Total steps: 147,660,800 Steps per second: 13,358 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1377.0 Actor loss: -9.5487 Critic loss: +0.0067 World model loss: +0.5877 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0553 Episode_Reward/track_angular_velocity: +52.1288 Episode_Reward/upright: +26.4875 Episode_Reward/leg_posture: -2.5502 Episode_Reward/standstill: -6.4973 Episode_Reward/action_rate_l2: -0.5596 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0819 Episode_Reward/rolling_leg_motion: -0.5869 Episode_Reward/obstacle_progress: +3.8609 Episode_Reward/wheel_body_intrusion: -0.2630 Episode_Reward/pair_body_intrusion: -0.2546 Curriculum/terrain_level_mean: +4.0825 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4870 Episode_Termination/fell_over: 0.0195 Episode_Termination/course_exit: 0.4740 Episode_Termination/illegal_nonwheel_contact: 0.0260 ============================================================================== ============================================================================== Learning iteration 36100/50000 Total steps: 147,865,600 Steps per second: 13,361 Iteration time: 0.307s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1314.5 Actor loss: -9.4899 Critic loss: +0.0047 World model loss: +0.5932 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6802 Episode_Reward/track_angular_velocity: +49.6462 Episode_Reward/upright: +25.2078 Episode_Reward/leg_posture: -2.5019 Episode_Reward/standstill: -6.3410 Episode_Reward/action_rate_l2: -0.5414 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0180 Episode_Reward/unnecessary_wheel_lift: -0.0722 Episode_Reward/rolling_leg_motion: -0.5591 Episode_Reward/obstacle_progress: +3.7167 Episode_Reward/wheel_body_intrusion: -0.2591 Episode_Reward/pair_body_intrusion: -0.2779 Curriculum/terrain_level_mean: +4.0842 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4276 Episode_Termination/fell_over: 0.0263 Episode_Termination/course_exit: 0.5263 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[589] ============================================================================== Learning iteration 36150/50000 Total steps: 148,070,400 Steps per second: 13,364 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1303.5 Actor loss: -9.5202 Critic loss: +0.0044 World model loss: +0.5705 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2099 Episode_Reward/track_angular_velocity: +49.2011 Episode_Reward/upright: +24.9899 Episode_Reward/leg_posture: -2.5192 Episode_Reward/standstill: -6.6538 Episode_Reward/action_rate_l2: -0.5306 Episode_Reward/joint_torques_l2: -0.1426 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0255 Episode_Reward/unnecessary_wheel_lift: -0.0833 Episode_Reward/rolling_leg_motion: -0.5514 Episode_Reward/obstacle_progress: +3.6719 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.2681 Curriculum/terrain_level_mean: +4.0869 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0248 Episode_Termination/course_exit: 0.5093 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== ============================================================================== Learning iteration 36200/50000 Total steps: 148,275,200 Steps per second: 13,367 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1340.1 Actor loss: -9.4880 Critic loss: +0.0060 World model loss: +0.5877 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7217 Episode_Reward/track_angular_velocity: +50.6391 Episode_Reward/upright: +25.6936 Episode_Reward/leg_posture: -2.5480 Episode_Reward/standstill: -6.5632 Episode_Reward/action_rate_l2: -0.5403 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0340 Episode_Reward/unnecessary_wheel_lift: -0.0898 Episode_Reward/rolling_leg_motion: -0.5688 Episode_Reward/obstacle_progress: +3.7266 Episode_Reward/wheel_body_intrusion: -0.2820 Episode_Reward/pair_body_intrusion: -0.3270 Curriculum/terrain_level_mean: +4.0867 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4744 Episode_Termination/fell_over: 0.0641 Episode_Termination/course_exit: 0.4423 Episode_Termination/illegal_nonwheel_contact: 0.0256 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 36250/50000 Total steps: 148,480,000 Steps per second: 13,370 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1292.4 Actor loss: -9.5572 Critic loss: +0.0035 World model loss: +0.5697 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9847 Episode_Reward/track_angular_velocity: +48.8521 Episode_Reward/upright: +24.7951 Episode_Reward/leg_posture: -2.4658 Episode_Reward/standstill: -6.3330 Episode_Reward/action_rate_l2: -0.5350 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0177 Episode_Reward/unnecessary_wheel_lift: -0.0682 Episode_Reward/rolling_leg_motion: -0.5518 Episode_Reward/obstacle_progress: +3.6083 Episode_Reward/wheel_body_intrusion: -0.2519 Episode_Reward/pair_body_intrusion: -0.2940 Curriculum/terrain_level_mean: +4.0884 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.3916 Episode_Termination/fell_over: 0.0210 Episode_Termination/course_exit: 0.5594 Episode_Termination/illegal_nonwheel_contact: 0.0280 ============================================================================== ============================================================================== Learning iteration 36300/50000 Total steps: 148,684,800 Steps per second: 13,372 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1380.2 Actor loss: -9.5247 Critic loss: +0.0041 World model loss: +0.5857 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2812 Episode_Reward/track_angular_velocity: +52.1969 Episode_Reward/upright: +26.4907 Episode_Reward/leg_posture: -2.6365 Episode_Reward/standstill: -6.2872 Episode_Reward/action_rate_l2: -0.5721 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0851 Episode_Reward/rolling_leg_motion: -0.5688 Episode_Reward/obstacle_progress: +3.8977 Episode_Reward/wheel_body_intrusion: -0.2813 Episode_Reward/pair_body_intrusion: -0.3312 Curriculum/terrain_level_mean: +4.0889 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4880 Episode_Termination/fell_over: 0.0480 Episode_Termination/course_exit: 0.4640 Episode_Termination/illegal_nonwheel_contact: 0.0080 ============================================================================== ============================================================================== Learning iteration 36350/50000 Total steps: 148,889,600 Steps per second: 13,375 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0875 Mean episode length: 1302.0 Actor loss: -9.5388 Critic loss: +0.0055 World model loss: +0.5726 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4299 Episode_Reward/track_angular_velocity: +49.2878 Episode_Reward/upright: +24.9648 Episode_Reward/leg_posture: -2.4424 Episode_Reward/standstill: -6.1504 Episode_Reward/action_rate_l2: -0.5307 Episode_Reward/joint_torques_l2: -0.1393 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0203 Episode_Reward/unnecessary_wheel_lift: -0.0828 Episode_Reward/rolling_leg_motion: -0.5456 Episode_Reward/obstacle_progress: +3.6547 Episode_Reward/wheel_body_intrusion: -0.2523 Episode_Reward/pair_body_intrusion: -0.2897 Curriculum/terrain_level_mean: +4.0906 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4451 Episode_Termination/fell_over: 0.0366 Episode_Termination/course_exit: 0.5061 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 36400/50000 Total steps: 149,094,400 Steps per second: 13,377 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1337.1 Actor loss: -9.5318 Critic loss: +0.0044 World model loss: +0.5922 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6876 Episode_Reward/track_angular_velocity: +50.5772 Episode_Reward/upright: +25.6484 Episode_Reward/leg_posture: -2.5541 Episode_Reward/standstill: -5.9911 Episode_Reward/action_rate_l2: -0.5436 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0261 Episode_Reward/unnecessary_wheel_lift: -0.0787 Episode_Reward/rolling_leg_motion: -0.5650 Episode_Reward/obstacle_progress: +3.7322 Episode_Reward/wheel_body_intrusion: -0.2553 Episode_Reward/pair_body_intrusion: -0.2666 Curriculum/terrain_level_mean: +4.0913 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0907 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4143 Episode_Termination/fell_over: 0.0429 Episode_Termination/course_exit: 0.5143 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1809] ============================================================================== Learning iteration 36450/50000 Total steps: 149,299,200 Steps per second: 13,380 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1333.3 Actor loss: -9.5288 Critic loss: +0.0041 World model loss: +0.5782 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4989 Episode_Reward/track_angular_velocity: +50.3841 Episode_Reward/upright: +25.5876 Episode_Reward/leg_posture: -2.5172 Episode_Reward/standstill: -6.4632 Episode_Reward/action_rate_l2: -0.5512 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0156 Episode_Reward/unnecessary_wheel_lift: -0.0915 Episode_Reward/rolling_leg_motion: -0.5755 Episode_Reward/obstacle_progress: +3.8021 Episode_Reward/wheel_body_intrusion: -0.2537 Episode_Reward/pair_body_intrusion: -0.2608 Curriculum/terrain_level_mean: +4.0935 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4277 Episode_Termination/fell_over: 0.0063 Episode_Termination/course_exit: 0.5346 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 36500/50000 Total steps: 149,504,000 Steps per second: 13,383 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1327.4 Actor loss: -9.5448 Critic loss: +0.0056 World model loss: +0.5845 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1298 Episode_Reward/track_angular_velocity: +50.2005 Episode_Reward/upright: +25.4514 Episode_Reward/leg_posture: -2.5666 Episode_Reward/standstill: -6.5691 Episode_Reward/action_rate_l2: -0.5348 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.0923 Episode_Reward/rolling_leg_motion: -0.5751 Episode_Reward/obstacle_progress: +3.7170 Episode_Reward/wheel_body_intrusion: -0.3016 Episode_Reward/pair_body_intrusion: -0.3001 Curriculum/terrain_level_mean: +4.0933 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4315 Episode_Termination/fell_over: 0.0411 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2153] ============================================================================== Learning iteration 36550/50000 Total steps: 149,708,800 Steps per second: 13,386 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1399.0 Actor loss: -9.5178 Critic loss: +0.0067 World model loss: +0.5885 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7934 Episode_Reward/track_angular_velocity: +52.9557 Episode_Reward/upright: +26.8417 Episode_Reward/leg_posture: -2.6765 Episode_Reward/standstill: -6.4392 Episode_Reward/action_rate_l2: -0.5730 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0936 Episode_Reward/rolling_leg_motion: -0.5932 Episode_Reward/obstacle_progress: +3.9486 Episode_Reward/wheel_body_intrusion: -0.3006 Episode_Reward/pair_body_intrusion: -0.3260 Curriculum/terrain_level_mean: +4.0955 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1324 Episode_Termination/time_out: 0.4966 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.4698 Episode_Termination/illegal_nonwheel_contact: 0.0201 ============================================================================== ============================================================================== Learning iteration 36600/50000 Total steps: 149,913,600 Steps per second: 13,388 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1331.9 Actor loss: -9.5356 Critic loss: +0.0048 World model loss: +0.6056 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2746 Episode_Reward/track_angular_velocity: +50.4116 Episode_Reward/upright: +25.5235 Episode_Reward/leg_posture: -2.5293 Episode_Reward/standstill: -6.3869 Episode_Reward/action_rate_l2: -0.5393 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0104 Episode_Reward/unnecessary_wheel_lift: -0.0911 Episode_Reward/rolling_leg_motion: -0.5446 Episode_Reward/obstacle_progress: +3.7678 Episode_Reward/wheel_body_intrusion: -0.2793 Episode_Reward/pair_body_intrusion: -0.3337 Curriculum/terrain_level_mean: +4.0969 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4219 Episode_Termination/fell_over: 0.0156 Episode_Termination/course_exit: 0.5469 Episode_Termination/illegal_nonwheel_contact: 0.0156 ============================================================================== ============================================================================== Learning iteration 36650/50000 Total steps: 150,118,400 Steps per second: 13,391 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1332.6 Actor loss: -9.5548 Critic loss: +0.0050 World model loss: +0.5755 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1063 Episode_Reward/track_angular_velocity: +50.3778 Episode_Reward/upright: +25.5410 Episode_Reward/leg_posture: -2.5390 Episode_Reward/standstill: -6.3998 Episode_Reward/action_rate_l2: -0.5437 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.0876 Episode_Reward/rolling_leg_motion: -0.5563 Episode_Reward/obstacle_progress: +3.6683 Episode_Reward/wheel_body_intrusion: -0.2667 Episode_Reward/pair_body_intrusion: -0.2901 Curriculum/terrain_level_mean: +4.0989 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4270 Episode_Termination/fell_over: 0.0337 Episode_Termination/course_exit: 0.5169 Episode_Termination/illegal_nonwheel_contact: 0.0393 ============================================================================== ============================================================================== Learning iteration 36700/50000 Total steps: 150,323,200 Steps per second: 13,394 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1345.5 Actor loss: -9.5162 Critic loss: +0.0135 World model loss: +0.5950 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6999 Episode_Reward/track_angular_velocity: +50.8880 Episode_Reward/upright: +25.7902 Episode_Reward/leg_posture: -2.5420 Episode_Reward/standstill: -6.3562 Episode_Reward/action_rate_l2: -0.5662 Episode_Reward/joint_torques_l2: -0.1455 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0864 Episode_Reward/rolling_leg_motion: -0.5762 Episode_Reward/obstacle_progress: +3.7469 Episode_Reward/wheel_body_intrusion: -0.2724 Episode_Reward/pair_body_intrusion: -0.2610 Curriculum/terrain_level_mean: +4.1003 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4503 Episode_Termination/fell_over: 0.0265 Episode_Termination/course_exit: 0.5033 Episode_Termination/illegal_nonwheel_contact: 0.0397 ============================================================================== ============================================================================== Learning iteration 36750/50000 Total steps: 150,528,000 Steps per second: 13,396 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1411.1 Actor loss: -9.5343 Critic loss: +0.0051 World model loss: +0.5649 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0178 Episode_Reward/track_angular_velocity: +53.3743 Episode_Reward/upright: +27.0369 Episode_Reward/leg_posture: -2.6583 Episode_Reward/standstill: -6.5215 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1524 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0083 Episode_Reward/unnecessary_wheel_lift: -0.0799 Episode_Reward/rolling_leg_motion: -0.5921 Episode_Reward/obstacle_progress: +3.9454 Episode_Reward/wheel_body_intrusion: -0.2866 Episode_Reward/pair_body_intrusion: -0.2760 Curriculum/terrain_level_mean: +4.1023 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5208 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4583 Episode_Termination/illegal_nonwheel_contact: 0.0208 ============================================================================== ============================================================================== Learning iteration 36800/50000 Total steps: 150,732,800 Steps per second: 13,399 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0872 Mean episode length: 1384.3 Actor loss: -9.5498 Critic loss: +0.0128 World model loss: +0.6132 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1270 Episode_Reward/track_angular_velocity: +52.3256 Episode_Reward/upright: +26.4982 Episode_Reward/leg_posture: -2.6551 Episode_Reward/standstill: -6.4292 Episode_Reward/action_rate_l2: -0.5758 Episode_Reward/joint_torques_l2: -0.1506 Episode_Reward/dof_pos_limits: -0.0017 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0693 Episode_Reward/rolling_leg_motion: -0.5811 Episode_Reward/obstacle_progress: +3.8267 Episode_Reward/wheel_body_intrusion: -0.2910 Episode_Reward/pair_body_intrusion: -0.3035 Curriculum/terrain_level_mean: +4.1045 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4741 Episode_Termination/fell_over: 0.0148 Episode_Termination/course_exit: 0.4963 Episode_Termination/illegal_nonwheel_contact: 0.0148 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2852] ============================================================================== Learning iteration 36850/50000 Total steps: 150,937,600 Steps per second: 13,402 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0868 Mean episode length: 1324.0 Actor loss: -9.5336 Critic loss: +0.0045 World model loss: +0.5702 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9973 Episode_Reward/track_angular_velocity: +50.0300 Episode_Reward/upright: +25.3360 Episode_Reward/leg_posture: -2.5455 Episode_Reward/standstill: -6.6099 Episode_Reward/action_rate_l2: -0.5439 Episode_Reward/joint_torques_l2: -0.1446 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0310 Episode_Reward/unnecessary_wheel_lift: -0.0990 Episode_Reward/rolling_leg_motion: -0.5637 Episode_Reward/obstacle_progress: +3.6455 Episode_Reward/wheel_body_intrusion: -0.2855 Episode_Reward/pair_body_intrusion: -0.2606 Curriculum/terrain_level_mean: +4.1045 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.3935 Episode_Termination/fell_over: 0.0387 Episode_Termination/course_exit: 0.5290 Episode_Termination/illegal_nonwheel_contact: 0.0581 ============================================================================== ============================================================================== Learning iteration 36900/50000 Total steps: 151,142,400 Steps per second: 13,404 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1321.9 Actor loss: -9.5271 Critic loss: +0.0049 World model loss: +0.5888 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0339 Episode_Reward/track_angular_velocity: +49.9191 Episode_Reward/upright: +25.3121 Episode_Reward/leg_posture: -2.5196 Episode_Reward/standstill: -6.3600 Episode_Reward/action_rate_l2: -0.5584 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0232 Episode_Reward/unnecessary_wheel_lift: -0.0904 Episode_Reward/rolling_leg_motion: -0.5628 Episode_Reward/obstacle_progress: +3.6481 Episode_Reward/wheel_body_intrusion: -0.2640 Episode_Reward/pair_body_intrusion: -0.2447 Curriculum/terrain_level_mean: +4.1060 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4233 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.5215 Episode_Termination/illegal_nonwheel_contact: 0.0491 ============================================================================== ============================================================================== Learning iteration 36950/50000 Total steps: 151,347,200 Steps per second: 13,407 Iteration time: 0.306s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1435.3 Actor loss: -9.5474 Critic loss: +0.0064 World model loss: +0.6162 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +51.2485 Episode_Reward/track_angular_velocity: +54.3216 Episode_Reward/upright: +27.5021 Episode_Reward/leg_posture: -2.7717 Episode_Reward/standstill: -6.9425 Episode_Reward/action_rate_l2: -0.5933 Episode_Reward/joint_torques_l2: -0.1545 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0168 Episode_Reward/unnecessary_wheel_lift: -0.0879 Episode_Reward/rolling_leg_motion: -0.6249 Episode_Reward/obstacle_progress: +3.9391 Episode_Reward/wheel_body_intrusion: -0.2888 Episode_Reward/pair_body_intrusion: -0.2968 Curriculum/terrain_level_mean: +4.1079 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0900 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.5163 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0196 ============================================================================== ============================================================================== Learning iteration 37000/50000 Total steps: 151,552,000 Steps per second: 13,409 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1324.8 Actor loss: -9.5351 Critic loss: +0.0040 World model loss: +0.5947 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3377 Episode_Reward/track_angular_velocity: +50.1359 Episode_Reward/upright: +25.3866 Episode_Reward/leg_posture: -2.5269 Episode_Reward/standstill: -6.2430 Episode_Reward/action_rate_l2: -0.5453 Episode_Reward/joint_torques_l2: -0.1432 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0829 Episode_Reward/rolling_leg_motion: -0.5779 Episode_Reward/obstacle_progress: +3.6288 Episode_Reward/wheel_body_intrusion: -0.2715 Episode_Reward/pair_body_intrusion: -0.3084 Curriculum/terrain_level_mean: +4.1096 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0897 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4595 Episode_Termination/fell_over: 0.0135 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0338 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2380] ============================================================================== Learning iteration 37050/50000 Total steps: 151,756,800 Steps per second: 13,412 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1402.3 Actor loss: -9.5067 Critic loss: +0.0056 World model loss: +0.6025 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0455 Episode_Reward/track_angular_velocity: +53.0687 Episode_Reward/upright: +26.8403 Episode_Reward/leg_posture: -2.6885 Episode_Reward/standstill: -6.6570 Episode_Reward/action_rate_l2: -0.5806 Episode_Reward/joint_torques_l2: -0.1522 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0192 Episode_Reward/unnecessary_wheel_lift: -0.0809 Episode_Reward/rolling_leg_motion: -0.5969 Episode_Reward/obstacle_progress: +3.9076 Episode_Reward/wheel_body_intrusion: -0.2903 Episode_Reward/pair_body_intrusion: -0.3373 Curriculum/terrain_level_mean: +4.1104 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0903 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.4895 Episode_Termination/fell_over: 0.0140 Episode_Termination/course_exit: 0.4615 Episode_Termination/illegal_nonwheel_contact: 0.0420 ============================================================================== ============================================================================== Learning iteration 37100/50000 Total steps: 151,961,600 Steps per second: 13,415 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0867 Mean episode length: 1368.8 Actor loss: -9.5380 Critic loss: +0.0038 World model loss: +0.5752 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6767 Episode_Reward/track_angular_velocity: +51.7820 Episode_Reward/upright: +26.1929 Episode_Reward/leg_posture: -2.6337 Episode_Reward/standstill: -7.2854 Episode_Reward/action_rate_l2: -0.5789 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0838 Episode_Reward/rolling_leg_motion: -0.5758 Episode_Reward/obstacle_progress: +3.8480 Episode_Reward/wheel_body_intrusion: -0.2711 Episode_Reward/pair_body_intrusion: -0.2929 Curriculum/terrain_level_mean: +4.1111 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4779 Episode_Termination/fell_over: 0.0074 Episode_Termination/course_exit: 0.4779 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 37150/50000 Total steps: 152,166,400 Steps per second: 13,417 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1368.7 Actor loss: -9.5096 Critic loss: +0.0045 World model loss: +0.5910 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5233 Episode_Reward/track_angular_velocity: +51.7205 Episode_Reward/upright: +26.1680 Episode_Reward/leg_posture: -2.6480 Episode_Reward/standstill: -6.8385 Episode_Reward/action_rate_l2: -0.5835 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.0827 Episode_Reward/rolling_leg_motion: -0.5776 Episode_Reward/obstacle_progress: +3.7718 Episode_Reward/wheel_body_intrusion: -0.2878 Episode_Reward/pair_body_intrusion: -0.2861 Curriculum/terrain_level_mean: +4.1108 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4795 Episode_Termination/fell_over: 0.0479 Episode_Termination/course_exit: 0.4521 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 37200/50000 Total steps: 152,371,200 Steps per second: 13,420 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1406.1 Actor loss: -9.5014 Critic loss: +0.0100 World model loss: +0.5973 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8344 Episode_Reward/track_angular_velocity: +53.0987 Episode_Reward/upright: +26.8741 Episode_Reward/leg_posture: -2.7520 Episode_Reward/standstill: -6.7152 Episode_Reward/action_rate_l2: -0.5954 Episode_Reward/joint_torques_l2: -0.1528 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.5911 Episode_Reward/obstacle_progress: +3.7879 Episode_Reward/wheel_body_intrusion: -0.3030 Episode_Reward/pair_body_intrusion: -0.3277 Curriculum/terrain_level_mean: +4.1116 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5649 Episode_Termination/fell_over: 0.0153 Episode_Termination/course_exit: 0.3893 Episode_Termination/illegal_nonwheel_contact: 0.0382 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1438] ============================================================================== Learning iteration 37250/50000 Total steps: 152,576,000 Steps per second: 13,422 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1365.5 Actor loss: -9.5161 Critic loss: +0.0037 World model loss: +0.5808 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2686 Episode_Reward/track_angular_velocity: +51.6016 Episode_Reward/upright: +26.1033 Episode_Reward/leg_posture: -2.6564 Episode_Reward/standstill: -6.4043 Episode_Reward/action_rate_l2: -0.5808 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0937 Episode_Reward/rolling_leg_motion: -0.5783 Episode_Reward/obstacle_progress: +3.6582 Episode_Reward/wheel_body_intrusion: -0.2827 Episode_Reward/pair_body_intrusion: -0.3147 Curriculum/terrain_level_mean: +4.1128 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5031 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.4591 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 37300/50000 Total steps: 152,780,800 Steps per second: 13,425 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1337.6 Actor loss: -9.4911 Critic loss: +0.0075 World model loss: +0.5460 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4244 Episode_Reward/track_angular_velocity: +50.6040 Episode_Reward/upright: +25.6094 Episode_Reward/leg_posture: -2.5467 Episode_Reward/standstill: -6.1043 Episode_Reward/action_rate_l2: -0.5672 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0763 Episode_Reward/rolling_leg_motion: -0.5691 Episode_Reward/obstacle_progress: +3.6425 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.2949 Curriculum/terrain_level_mean: +4.1145 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4510 Episode_Termination/fell_over: 0.0065 Episode_Termination/course_exit: 0.4967 Episode_Termination/illegal_nonwheel_contact: 0.0458 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2305] ============================================================================== Learning iteration 37350/50000 Total steps: 152,985,600 Steps per second: 13,428 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1360.6 Actor loss: -9.5448 Critic loss: +0.0039 World model loss: +0.5621 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4087 Episode_Reward/track_angular_velocity: +51.4621 Episode_Reward/upright: +26.0944 Episode_Reward/leg_posture: -2.5509 Episode_Reward/standstill: -6.1541 Episode_Reward/action_rate_l2: -0.5717 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0345 Episode_Reward/unnecessary_wheel_lift: -0.1146 Episode_Reward/rolling_leg_motion: -0.5974 Episode_Reward/obstacle_progress: +3.7452 Episode_Reward/wheel_body_intrusion: -0.2513 Episode_Reward/pair_body_intrusion: -0.3131 Curriculum/terrain_level_mean: +4.1152 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4535 Episode_Termination/fell_over: 0.0523 Episode_Termination/course_exit: 0.4709 Episode_Termination/illegal_nonwheel_contact: 0.0407 ============================================================================== ============================================================================== Learning iteration 37400/50000 Total steps: 153,190,400 Steps per second: 13,430 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0877 Mean episode length: 1350.4 Actor loss: -9.5346 Critic loss: +0.0054 World model loss: +0.5779 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1587 Episode_Reward/track_angular_velocity: +51.1013 Episode_Reward/upright: +25.9299 Episode_Reward/leg_posture: -2.5617 Episode_Reward/standstill: -6.0410 Episode_Reward/action_rate_l2: -0.5579 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0211 Episode_Reward/unnecessary_wheel_lift: -0.0951 Episode_Reward/rolling_leg_motion: -0.5818 Episode_Reward/obstacle_progress: +3.7679 Episode_Reward/wheel_body_intrusion: -0.2550 Episode_Reward/pair_body_intrusion: -0.3124 Curriculum/terrain_level_mean: +4.1167 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4777 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4713 Episode_Termination/illegal_nonwheel_contact: 0.0318 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2587] ============================================================================== Learning iteration 37450/50000 Total steps: 153,395,200 Steps per second: 13,433 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1464.4 Actor loss: -9.5541 Critic loss: +0.0037 World model loss: +0.5520 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +52.0115 Episode_Reward/track_angular_velocity: +55.4012 Episode_Reward/upright: +28.0354 Episode_Reward/leg_posture: -2.8417 Episode_Reward/standstill: -6.5636 Episode_Reward/action_rate_l2: -0.6062 Episode_Reward/joint_torques_l2: -0.1587 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0086 Episode_Reward/unnecessary_wheel_lift: -0.0793 Episode_Reward/rolling_leg_motion: -0.6340 Episode_Reward/obstacle_progress: +3.9873 Episode_Reward/wheel_body_intrusion: -0.3133 Episode_Reward/pair_body_intrusion: -0.3218 Curriculum/terrain_level_mean: +4.1179 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.6000 Episode_Termination/fell_over: 0.0143 Episode_Termination/course_exit: 0.3786 Episode_Termination/illegal_nonwheel_contact: 0.0143 ============================================================================== ============================================================================== Learning iteration 37500/50000 Total steps: 153,600,000 Steps per second: 13,435 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1337.5 Actor loss: -9.5148 Critic loss: +0.0047 World model loss: +0.5564 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5162 Episode_Reward/track_angular_velocity: +50.5776 Episode_Reward/upright: +25.5787 Episode_Reward/leg_posture: -2.5994 Episode_Reward/standstill: -6.2101 Episode_Reward/action_rate_l2: -0.5581 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.1039 Episode_Reward/rolling_leg_motion: -0.5812 Episode_Reward/obstacle_progress: +3.6536 Episode_Reward/wheel_body_intrusion: -0.2671 Episode_Reward/pair_body_intrusion: -0.2887 Curriculum/terrain_level_mean: +4.1177 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4605 Episode_Termination/fell_over: 0.0395 Episode_Termination/course_exit: 0.4605 Episode_Termination/illegal_nonwheel_contact: 0.0526 ============================================================================== ============================================================================== Learning iteration 37550/50000 Total steps: 153,804,800 Steps per second: 13,437 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0872 Mean episode length: 1308.3 Actor loss: -9.5183 Critic loss: +0.0065 World model loss: +0.5646 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5708 Episode_Reward/track_angular_velocity: +49.4427 Episode_Reward/upright: +25.0346 Episode_Reward/leg_posture: -2.5378 Episode_Reward/standstill: -6.0681 Episode_Reward/action_rate_l2: -0.5275 Episode_Reward/joint_torques_l2: -0.1412 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0204 Episode_Reward/unnecessary_wheel_lift: -0.0806 Episode_Reward/rolling_leg_motion: -0.5751 Episode_Reward/obstacle_progress: +3.5662 Episode_Reward/wheel_body_intrusion: -0.2585 Episode_Reward/pair_body_intrusion: -0.2660 Curriculum/terrain_level_mean: +4.1182 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4275 Episode_Termination/fell_over: 0.0153 Episode_Termination/course_exit: 0.5115 Episode_Termination/illegal_nonwheel_contact: 0.0534 ============================================================================== ============================================================================== Learning iteration 37600/50000 Total steps: 154,009,600 Steps per second: 13,440 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1334.3 Actor loss: -9.5503 Critic loss: +0.0040 World model loss: +0.5905 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5757 Episode_Reward/track_angular_velocity: +50.4737 Episode_Reward/upright: +25.5714 Episode_Reward/leg_posture: -2.5916 Episode_Reward/standstill: -6.0128 Episode_Reward/action_rate_l2: -0.5349 Episode_Reward/joint_torques_l2: -0.1426 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0287 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.5875 Episode_Reward/obstacle_progress: +3.6170 Episode_Reward/wheel_body_intrusion: -0.2690 Episode_Reward/pair_body_intrusion: -0.2791 Curriculum/terrain_level_mean: +4.1189 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.4096 Episode_Termination/fell_over: 0.0542 Episode_Termination/course_exit: 0.5241 Episode_Termination/illegal_nonwheel_contact: 0.0482 ============================================================================== ============================================================================== Learning iteration 37650/50000 Total steps: 154,214,400 Steps per second: 13,443 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1410.1 Actor loss: -9.5027 Critic loss: +0.0038 World model loss: +0.5777 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3336 Episode_Reward/track_angular_velocity: +53.3996 Episode_Reward/upright: +27.0665 Episode_Reward/leg_posture: -2.7318 Episode_Reward/standstill: -6.3670 Episode_Reward/action_rate_l2: -0.5678 Episode_Reward/joint_torques_l2: -0.1498 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0220 Episode_Reward/unnecessary_wheel_lift: -0.0859 Episode_Reward/rolling_leg_motion: -0.6221 Episode_Reward/obstacle_progress: +3.7844 Episode_Reward/wheel_body_intrusion: -0.2683 Episode_Reward/pair_body_intrusion: -0.3217 Curriculum/terrain_level_mean: +4.1196 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0899 Metrics/base_velocity/error_vel_yaw: 0.1319 Episode_Termination/time_out: 0.5677 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.3806 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== ============================================================================== Learning iteration 37700/50000 Total steps: 154,419,200 Steps per second: 13,445 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1345.8 Actor loss: -9.5171 Critic loss: +0.0036 World model loss: +0.5657 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9988 Episode_Reward/track_angular_velocity: +50.9508 Episode_Reward/upright: +25.8387 Episode_Reward/leg_posture: -2.6014 Episode_Reward/standstill: -6.0405 Episode_Reward/action_rate_l2: -0.5404 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0131 Episode_Reward/unnecessary_wheel_lift: -0.0784 Episode_Reward/rolling_leg_motion: -0.5850 Episode_Reward/obstacle_progress: +3.6138 Episode_Reward/wheel_body_intrusion: -0.2551 Episode_Reward/pair_body_intrusion: -0.3096 Curriculum/terrain_level_mean: +4.1221 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1325 Episode_Termination/time_out: 0.4777 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.4904 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 37750/50000 Total steps: 154,624,000 Steps per second: 13,448 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1352.0 Actor loss: -9.5487 Critic loss: +0.0060 World model loss: +0.5767 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2113 Episode_Reward/track_angular_velocity: +51.1581 Episode_Reward/upright: +25.9206 Episode_Reward/leg_posture: -2.6214 Episode_Reward/standstill: -6.0831 Episode_Reward/action_rate_l2: -0.5426 Episode_Reward/joint_torques_l2: -0.1446 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0844 Episode_Reward/rolling_leg_motion: -0.5881 Episode_Reward/obstacle_progress: +3.6913 Episode_Reward/wheel_body_intrusion: -0.2619 Episode_Reward/pair_body_intrusion: -0.2939 Curriculum/terrain_level_mean: +4.1240 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.4522 Episode_Termination/fell_over: 0.0191 Episode_Termination/course_exit: 0.5159 Episode_Termination/illegal_nonwheel_contact: 0.0191 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2281] ============================================================================== Learning iteration 37800/50000 Total steps: 154,828,800 Steps per second: 13,450 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1371.7 Actor loss: -9.5249 Critic loss: +0.0054 World model loss: +0.5517 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9377 Episode_Reward/track_angular_velocity: +51.9756 Episode_Reward/upright: +26.2654 Episode_Reward/leg_posture: -2.6370 Episode_Reward/standstill: -6.2137 Episode_Reward/action_rate_l2: -0.5544 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0892 Episode_Reward/rolling_leg_motion: -0.5981 Episode_Reward/obstacle_progress: +3.7516 Episode_Reward/wheel_body_intrusion: -0.2653 Episode_Reward/pair_body_intrusion: -0.3393 Curriculum/terrain_level_mean: +4.1265 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1315 Episode_Termination/time_out: 0.4720 Episode_Termination/fell_over: 0.0124 Episode_Termination/course_exit: 0.4907 Episode_Termination/illegal_nonwheel_contact: 0.0311 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_037800.mp4 ============================================================================== Learning iteration 37850/50000 Total steps: 155,033,600 Steps per second: 13,446 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1367.4 Actor loss: -9.5120 Critic loss: +0.0039 World model loss: +0.5686 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7949 Episode_Reward/track_angular_velocity: +51.7856 Episode_Reward/upright: +26.1836 Episode_Reward/leg_posture: -2.5914 Episode_Reward/standstill: -6.1007 Episode_Reward/action_rate_l2: -0.5572 Episode_Reward/joint_torques_l2: -0.1470 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0715 Episode_Reward/rolling_leg_motion: -0.5793 Episode_Reward/obstacle_progress: +3.8211 Episode_Reward/wheel_body_intrusion: -0.2496 Episode_Reward/pair_body_intrusion: -0.3087 Curriculum/terrain_level_mean: +4.1294 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.4825 Episode_Termination/fell_over: 0.0140 Episode_Termination/course_exit: 0.4965 Episode_Termination/illegal_nonwheel_contact: 0.0140 ============================================================================== ============================================================================== Learning iteration 37900/50000 Total steps: 155,238,400 Steps per second: 13,449 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1399.9 Actor loss: -9.5426 Critic loss: +0.0068 World model loss: +0.5803 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9304 Episode_Reward/track_angular_velocity: +52.9496 Episode_Reward/upright: +26.8131 Episode_Reward/leg_posture: -2.6952 Episode_Reward/standstill: -6.7299 Episode_Reward/action_rate_l2: -0.5785 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0191 Episode_Reward/unnecessary_wheel_lift: -0.1017 Episode_Reward/rolling_leg_motion: -0.6071 Episode_Reward/obstacle_progress: +3.8790 Episode_Reward/wheel_body_intrusion: -0.2825 Episode_Reward/pair_body_intrusion: -0.2904 Curriculum/terrain_level_mean: +4.1316 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0907 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4839 Episode_Termination/fell_over: 0.0387 Episode_Termination/course_exit: 0.4710 Episode_Termination/illegal_nonwheel_contact: 0.0129 ============================================================================== ============================================================================== Learning iteration 37950/50000 Total steps: 155,443,200 Steps per second: 13,451 Iteration time: 0.305s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1412.6 Actor loss: -9.5148 Critic loss: +0.0030 World model loss: +0.5582 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.3914 Episode_Reward/track_angular_velocity: +53.4162 Episode_Reward/upright: +27.1006 Episode_Reward/leg_posture: -2.7135 Episode_Reward/standstill: -6.6710 Episode_Reward/action_rate_l2: -0.5836 Episode_Reward/joint_torques_l2: -0.1509 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.6148 Episode_Reward/obstacle_progress: +3.8972 Episode_Reward/wheel_body_intrusion: -0.2925 Episode_Reward/pair_body_intrusion: -0.2932 Curriculum/terrain_level_mean: +4.1333 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5302 Episode_Termination/fell_over: 0.0134 Episode_Termination/course_exit: 0.4362 Episode_Termination/illegal_nonwheel_contact: 0.0268 ============================================================================== ============================================================================== Learning iteration 38000/50000 Total steps: 155,648,000 Steps per second: 13,454 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1341.1 Actor loss: -9.5422 Critic loss: +0.0045 World model loss: +0.5905 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8270 Episode_Reward/track_angular_velocity: +50.7480 Episode_Reward/upright: +25.7149 Episode_Reward/leg_posture: -2.6000 Episode_Reward/standstill: -6.3609 Episode_Reward/action_rate_l2: -0.5544 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.0728 Episode_Reward/rolling_leg_motion: -0.5635 Episode_Reward/obstacle_progress: +3.7613 Episode_Reward/wheel_body_intrusion: -0.2687 Episode_Reward/pair_body_intrusion: -0.2797 Curriculum/terrain_level_mean: +4.1355 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0907 Metrics/base_velocity/error_vel_yaw: 0.1329 Episode_Termination/time_out: 0.4348 Episode_Termination/fell_over: 0.0362 Episode_Termination/course_exit: 0.5072 Episode_Termination/illegal_nonwheel_contact: 0.0435 ============================================================================== ============================================================================== Learning iteration 38050/50000 Total steps: 155,852,800 Steps per second: 13,456 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1301.1 Actor loss: -9.5409 Critic loss: +0.0057 World model loss: +0.5791 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4460 Episode_Reward/track_angular_velocity: +49.2179 Episode_Reward/upright: +24.9484 Episode_Reward/leg_posture: -2.5456 Episode_Reward/standstill: -6.0230 Episode_Reward/action_rate_l2: -0.5503 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0203 Episode_Reward/unnecessary_wheel_lift: -0.0731 Episode_Reward/rolling_leg_motion: -0.5701 Episode_Reward/obstacle_progress: +3.5747 Episode_Reward/wheel_body_intrusion: -0.2732 Episode_Reward/pair_body_intrusion: -0.2775 Curriculum/terrain_level_mean: +4.1362 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0902 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4126 Episode_Termination/fell_over: 0.0210 Episode_Termination/course_exit: 0.5315 Episode_Termination/illegal_nonwheel_contact: 0.0420 ============================================================================== ============================================================================== Learning iteration 38100/50000 Total steps: 156,057,600 Steps per second: 13,459 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1407.2 Actor loss: -9.5180 Critic loss: +0.0037 World model loss: +0.5529 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2311 Episode_Reward/track_angular_velocity: +53.2496 Episode_Reward/upright: +27.0172 Episode_Reward/leg_posture: -2.6998 Episode_Reward/standstill: -6.6032 Episode_Reward/action_rate_l2: -0.5887 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0027 Episode_Reward/unnecessary_wheel_lift: -0.0808 Episode_Reward/rolling_leg_motion: -0.6188 Episode_Reward/obstacle_progress: +3.8689 Episode_Reward/wheel_body_intrusion: -0.2874 Episode_Reward/pair_body_intrusion: -0.3250 Curriculum/terrain_level_mean: +4.1392 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.5074 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4853 Episode_Termination/illegal_nonwheel_contact: 0.0074 ============================================================================== ============================================================================== Learning iteration 38150/50000 Total steps: 156,262,400 Steps per second: 13,461 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1371.5 Actor loss: -9.5210 Critic loss: +0.0033 World model loss: +0.5579 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8713 Episode_Reward/track_angular_velocity: +51.8592 Episode_Reward/upright: +26.3144 Episode_Reward/leg_posture: -2.6202 Episode_Reward/standstill: -6.3726 Episode_Reward/action_rate_l2: -0.5711 Episode_Reward/joint_torques_l2: -0.1476 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0747 Episode_Reward/rolling_leg_motion: -0.5890 Episode_Reward/obstacle_progress: +3.8342 Episode_Reward/wheel_body_intrusion: -0.2651 Episode_Reward/pair_body_intrusion: -0.3056 Curriculum/terrain_level_mean: +4.1389 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4844 Episode_Termination/fell_over: 0.0312 Episode_Termination/course_exit: 0.4609 Episode_Termination/illegal_nonwheel_contact: 0.0312 ============================================================================== ============================================================================== Learning iteration 38200/50000 Total steps: 156,467,200 Steps per second: 13,464 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1395.9 Actor loss: -9.5252 Critic loss: +0.0059 World model loss: +0.5635 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7345 Episode_Reward/track_angular_velocity: +52.7411 Episode_Reward/upright: +26.7701 Episode_Reward/leg_posture: -2.6734 Episode_Reward/standstill: -6.0539 Episode_Reward/action_rate_l2: -0.5822 Episode_Reward/joint_torques_l2: -0.1489 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.0710 Episode_Reward/rolling_leg_motion: -0.5858 Episode_Reward/obstacle_progress: +3.9030 Episode_Reward/wheel_body_intrusion: -0.2507 Episode_Reward/pair_body_intrusion: -0.2870 Curriculum/terrain_level_mean: +4.1396 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0910 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.4521 Episode_Termination/illegal_nonwheel_contact: 0.0548 ============================================================================== ============================================================================== Learning iteration 38250/50000 Total steps: 156,672,000 Steps per second: 13,466 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1310.0 Actor loss: -9.5293 Critic loss: +0.0037 World model loss: +0.5554 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7867 Episode_Reward/track_angular_velocity: +49.5353 Episode_Reward/upright: +25.1380 Episode_Reward/leg_posture: -2.5072 Episode_Reward/standstill: -5.5352 Episode_Reward/action_rate_l2: -0.5394 Episode_Reward/joint_torques_l2: -0.1394 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0179 Episode_Reward/unnecessary_wheel_lift: -0.0903 Episode_Reward/rolling_leg_motion: -0.5698 Episode_Reward/obstacle_progress: +3.6363 Episode_Reward/wheel_body_intrusion: -0.2478 Episode_Reward/pair_body_intrusion: -0.2462 Curriculum/terrain_level_mean: +4.1421 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0900 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4465 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.5094 Episode_Termination/illegal_nonwheel_contact: 0.0440 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1413] ============================================================================== Learning iteration 38300/50000 Total steps: 156,876,800 Steps per second: 13,469 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1364.1 Actor loss: -9.5380 Critic loss: +0.0039 World model loss: +0.5856 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6526 Episode_Reward/track_angular_velocity: +51.5783 Episode_Reward/upright: +26.1728 Episode_Reward/leg_posture: -2.6464 Episode_Reward/standstill: -5.9708 Episode_Reward/action_rate_l2: -0.5584 Episode_Reward/joint_torques_l2: -0.1467 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0186 Episode_Reward/unnecessary_wheel_lift: -0.0893 Episode_Reward/rolling_leg_motion: -0.5856 Episode_Reward/obstacle_progress: +3.8107 Episode_Reward/wheel_body_intrusion: -0.2681 Episode_Reward/pair_body_intrusion: -0.3325 Curriculum/terrain_level_mean: +4.1443 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4678 Episode_Termination/fell_over: 0.0234 Episode_Termination/course_exit: 0.4912 Episode_Termination/illegal_nonwheel_contact: 0.0234 ============================================================================== ============================================================================== Learning iteration 38350/50000 Total steps: 157,081,600 Steps per second: 13,471 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0876 Mean episode length: 1309.0 Actor loss: -9.5181 Critic loss: +0.0055 World model loss: +0.6332 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5703 Episode_Reward/track_angular_velocity: +49.4791 Episode_Reward/upright: +25.1257 Episode_Reward/leg_posture: -2.5286 Episode_Reward/standstill: -5.6469 Episode_Reward/action_rate_l2: -0.5481 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0153 Episode_Reward/unnecessary_wheel_lift: -0.0688 Episode_Reward/rolling_leg_motion: -0.5623 Episode_Reward/obstacle_progress: +3.6500 Episode_Reward/wheel_body_intrusion: -0.2623 Episode_Reward/pair_body_intrusion: -0.3208 Curriculum/terrain_level_mean: +4.1458 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4485 Episode_Termination/fell_over: 0.0368 Episode_Termination/course_exit: 0.5074 Episode_Termination/illegal_nonwheel_contact: 0.0147 ============================================================================== ============================================================================== Learning iteration 38400/50000 Total steps: 157,286,400 Steps per second: 13,473 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1428.4 Actor loss: -9.5159 Critic loss: +0.0045 World model loss: +0.5716 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.6348 Episode_Reward/track_angular_velocity: +53.8420 Episode_Reward/upright: +27.4090 Episode_Reward/leg_posture: -2.7083 Episode_Reward/standstill: -6.7127 Episode_Reward/action_rate_l2: -0.6076 Episode_Reward/joint_torques_l2: -0.1537 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0251 Episode_Reward/unnecessary_wheel_lift: -0.0955 Episode_Reward/rolling_leg_motion: -0.6193 Episode_Reward/obstacle_progress: +4.0052 Episode_Reward/wheel_body_intrusion: -0.3017 Episode_Reward/pair_body_intrusion: -0.3522 Curriculum/terrain_level_mean: +4.1462 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1373 Episode_Termination/time_out: 0.5503 Episode_Termination/fell_over: 0.0355 Episode_Termination/course_exit: 0.3964 Episode_Termination/illegal_nonwheel_contact: 0.0296 ============================================================================== ============================================================================== Learning iteration 38450/50000 Total steps: 157,491,200 Steps per second: 13,476 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1361.2 Actor loss: -9.5343 Critic loss: +0.0059 World model loss: +0.6015 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4793 Episode_Reward/track_angular_velocity: +51.4447 Episode_Reward/upright: +26.1824 Episode_Reward/leg_posture: -2.5306 Episode_Reward/standstill: -5.8178 Episode_Reward/action_rate_l2: -0.5659 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0292 Episode_Reward/unnecessary_wheel_lift: -0.0916 Episode_Reward/rolling_leg_motion: -0.5869 Episode_Reward/obstacle_progress: +3.8413 Episode_Reward/wheel_body_intrusion: -0.2632 Episode_Reward/pair_body_intrusion: -0.2841 Curriculum/terrain_level_mean: +4.1484 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4841 Episode_Termination/fell_over: 0.0382 Episode_Termination/course_exit: 0.4459 Episode_Termination/illegal_nonwheel_contact: 0.0446 ============================================================================== ============================================================================== Learning iteration 38500/50000 Total steps: 157,696,000 Steps per second: 13,478 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0879 Mean episode length: 1404.2 Actor loss: -9.5509 Critic loss: +0.0033 World model loss: +0.5531 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1338 Episode_Reward/track_angular_velocity: +53.0728 Episode_Reward/upright: +26.9999 Episode_Reward/leg_posture: -2.6610 Episode_Reward/standstill: -6.1788 Episode_Reward/action_rate_l2: -0.5884 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.5992 Episode_Reward/obstacle_progress: +4.0694 Episode_Reward/wheel_body_intrusion: -0.2943 Episode_Reward/pair_body_intrusion: -0.3000 Curriculum/terrain_level_mean: +4.1504 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0904 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4843 Episode_Termination/fell_over: 0.0126 Episode_Termination/course_exit: 0.4843 Episode_Termination/illegal_nonwheel_contact: 0.0252 ============================================================================== ============================================================================== Learning iteration 38550/50000 Total steps: 157,900,800 Steps per second: 13,481 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1322.9 Actor loss: -9.5193 Critic loss: +0.0063 World model loss: +0.5845 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0741 Episode_Reward/track_angular_velocity: +50.0501 Episode_Reward/upright: +25.4260 Episode_Reward/leg_posture: -2.5149 Episode_Reward/standstill: -6.2336 Episode_Reward/action_rate_l2: -0.5602 Episode_Reward/joint_torques_l2: -0.1404 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.0761 Episode_Reward/rolling_leg_motion: -0.5573 Episode_Reward/obstacle_progress: +3.7226 Episode_Reward/wheel_body_intrusion: -0.2840 Episode_Reward/pair_body_intrusion: -0.3039 Curriculum/terrain_level_mean: +4.1509 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.4113 Episode_Termination/fell_over: 0.0284 Episode_Termination/course_exit: 0.5106 Episode_Termination/illegal_nonwheel_contact: 0.0567 ============================================================================== ============================================================================== Learning iteration 38600/50000 Total steps: 158,105,600 Steps per second: 13,483 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1356.7 Actor loss: -9.5095 Critic loss: +0.0073 World model loss: +0.5369 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3608 Episode_Reward/track_angular_velocity: +51.3515 Episode_Reward/upright: +26.1011 Episode_Reward/leg_posture: -2.5636 Episode_Reward/standstill: -6.1662 Episode_Reward/action_rate_l2: -0.5738 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0737 Episode_Reward/rolling_leg_motion: -0.5807 Episode_Reward/obstacle_progress: +3.8327 Episode_Reward/wheel_body_intrusion: -0.2883 Episode_Reward/pair_body_intrusion: -0.3061 Curriculum/terrain_level_mean: +4.1514 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1327 Episode_Termination/time_out: 0.4813 Episode_Termination/fell_over: 0.0250 Episode_Termination/course_exit: 0.4750 Episode_Termination/illegal_nonwheel_contact: 0.0312 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2352] RECOVERY reset 1 invalid environments; first IDs=[625] ============================================================================== Learning iteration 38650/50000 Total steps: 158,310,400 Steps per second: 13,486 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1327.8 Actor loss: -9.5336 Critic loss: +0.0047 World model loss: +0.5606 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3114 Episode_Reward/track_angular_velocity: +50.2529 Episode_Reward/upright: +25.5122 Episode_Reward/leg_posture: -2.5381 Episode_Reward/standstill: -5.9511 Episode_Reward/action_rate_l2: -0.5665 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0259 Episode_Reward/unnecessary_wheel_lift: -0.0912 Episode_Reward/rolling_leg_motion: -0.5856 Episode_Reward/obstacle_progress: +3.7618 Episode_Reward/wheel_body_intrusion: -0.2809 Episode_Reward/pair_body_intrusion: -0.3217 Curriculum/terrain_level_mean: +4.1516 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4233 Episode_Termination/fell_over: 0.0368 Episode_Termination/course_exit: 0.5215 Episode_Termination/illegal_nonwheel_contact: 0.0245 ============================================================================== ============================================================================== Learning iteration 38700/50000 Total steps: 158,515,200 Steps per second: 13,488 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1394.5 Actor loss: -9.5461 Critic loss: +0.0057 World model loss: +0.5710 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6370 Episode_Reward/track_angular_velocity: +52.7633 Episode_Reward/upright: +26.7828 Episode_Reward/leg_posture: -2.6323 Episode_Reward/standstill: -6.2874 Episode_Reward/action_rate_l2: -0.5870 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0191 Episode_Reward/unnecessary_wheel_lift: -0.0928 Episode_Reward/rolling_leg_motion: -0.6107 Episode_Reward/obstacle_progress: +3.9182 Episode_Reward/wheel_body_intrusion: -0.2762 Episode_Reward/pair_body_intrusion: -0.3216 Curriculum/terrain_level_mean: +4.1526 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1331 Episode_Termination/time_out: 0.5109 Episode_Termination/fell_over: 0.0292 Episode_Termination/course_exit: 0.4380 Episode_Termination/illegal_nonwheel_contact: 0.0365 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[455] ============================================================================== Learning iteration 38750/50000 Total steps: 158,720,000 Steps per second: 13,491 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1373.0 Actor loss: -9.4837 Critic loss: +0.0047 World model loss: +0.5867 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8395 Episode_Reward/track_angular_velocity: +51.9395 Episode_Reward/upright: +26.4030 Episode_Reward/leg_posture: -2.5999 Episode_Reward/standstill: -6.2099 Episode_Reward/action_rate_l2: -0.5755 Episode_Reward/joint_torques_l2: -0.1457 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0869 Episode_Reward/rolling_leg_motion: -0.5886 Episode_Reward/obstacle_progress: +3.8510 Episode_Reward/wheel_body_intrusion: -0.2881 Episode_Reward/pair_body_intrusion: -0.3182 Curriculum/terrain_level_mean: +4.1550 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4658 Episode_Termination/fell_over: 0.0137 Episode_Termination/course_exit: 0.5137 Episode_Termination/illegal_nonwheel_contact: 0.0205 ============================================================================== ============================================================================== Learning iteration 38800/50000 Total steps: 158,924,800 Steps per second: 13,493 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1332.1 Actor loss: -9.5400 Critic loss: +0.0043 World model loss: +0.5993 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2258 Episode_Reward/track_angular_velocity: +50.3632 Episode_Reward/upright: +25.5387 Episode_Reward/leg_posture: -2.5299 Episode_Reward/standstill: -6.4826 Episode_Reward/action_rate_l2: -0.5675 Episode_Reward/joint_torques_l2: -0.1422 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.0844 Episode_Reward/rolling_leg_motion: -0.5684 Episode_Reward/obstacle_progress: +3.7484 Episode_Reward/wheel_body_intrusion: -0.2873 Episode_Reward/pair_body_intrusion: -0.2948 Curriculum/terrain_level_mean: +4.1567 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4459 Episode_Termination/fell_over: 0.0338 Episode_Termination/course_exit: 0.5203 Episode_Termination/illegal_nonwheel_contact: 0.0203 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1379] ============================================================================== Learning iteration 38850/50000 Total steps: 159,129,600 Steps per second: 13,496 Iteration time: 0.304s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1344.4 Actor loss: -9.5560 Critic loss: +0.0036 World model loss: +0.5498 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5788 Episode_Reward/track_angular_velocity: +50.8573 Episode_Reward/upright: +25.7690 Episode_Reward/leg_posture: -2.5519 Episode_Reward/standstill: -6.2349 Episode_Reward/action_rate_l2: -0.5673 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0236 Episode_Reward/unnecessary_wheel_lift: -0.0966 Episode_Reward/rolling_leg_motion: -0.5735 Episode_Reward/obstacle_progress: +3.8307 Episode_Reward/wheel_body_intrusion: -0.2767 Episode_Reward/pair_body_intrusion: -0.2906 Curriculum/terrain_level_mean: +4.1580 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4634 Episode_Termination/fell_over: 0.0366 Episode_Termination/course_exit: 0.4817 Episode_Termination/illegal_nonwheel_contact: 0.0366 ============================================================================== ============================================================================== Learning iteration 38900/50000 Total steps: 159,334,400 Steps per second: 13,498 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1347.7 Actor loss: -9.5236 Critic loss: +0.0065 World model loss: +0.5754 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6291 Episode_Reward/track_angular_velocity: +50.9295 Episode_Reward/upright: +25.8492 Episode_Reward/leg_posture: -2.5459 Episode_Reward/standstill: -5.9746 Episode_Reward/action_rate_l2: -0.5755 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0800 Episode_Reward/rolling_leg_motion: -0.5694 Episode_Reward/obstacle_progress: +3.7924 Episode_Reward/wheel_body_intrusion: -0.2796 Episode_Reward/pair_body_intrusion: -0.2906 Curriculum/terrain_level_mean: +4.1589 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4467 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.5067 Episode_Termination/illegal_nonwheel_contact: 0.0333 ============================================================================== ============================================================================== Learning iteration 38950/50000 Total steps: 159,539,200 Steps per second: 13,501 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1358.2 Actor loss: -9.5195 Critic loss: +0.0051 World model loss: +0.5734 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0194 Episode_Reward/track_angular_velocity: +51.2879 Episode_Reward/upright: +26.0444 Episode_Reward/leg_posture: -2.5494 Episode_Reward/standstill: -5.9343 Episode_Reward/action_rate_l2: -0.5727 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0186 Episode_Reward/unnecessary_wheel_lift: -0.0729 Episode_Reward/rolling_leg_motion: -0.5452 Episode_Reward/obstacle_progress: +3.8739 Episode_Reward/wheel_body_intrusion: -0.2883 Episode_Reward/pair_body_intrusion: -0.3457 Curriculum/terrain_level_mean: +4.1602 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4962 Episode_Termination/fell_over: 0.0153 Episode_Termination/course_exit: 0.4504 Episode_Termination/illegal_nonwheel_contact: 0.0458 ============================================================================== ============================================================================== Learning iteration 39000/50000 Total steps: 159,744,000 Steps per second: 13,503 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0873 Mean episode length: 1366.5 Actor loss: -9.5069 Critic loss: +0.0084 World model loss: +0.5582 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2435 Episode_Reward/track_angular_velocity: +51.5706 Episode_Reward/upright: +26.1675 Episode_Reward/leg_posture: -2.5752 Episode_Reward/standstill: -5.9906 Episode_Reward/action_rate_l2: -0.5876 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0267 Episode_Reward/unnecessary_wheel_lift: -0.0741 Episode_Reward/rolling_leg_motion: -0.5601 Episode_Reward/obstacle_progress: +3.8207 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.3184 Curriculum/terrain_level_mean: +4.1604 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0949 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4551 Episode_Termination/fell_over: 0.0449 Episode_Termination/course_exit: 0.4808 Episode_Termination/illegal_nonwheel_contact: 0.0385 ============================================================================== ============================================================================== Learning iteration 39050/50000 Total steps: 159,948,800 Steps per second: 13,505 Iteration time: 0.303s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1314.9 Actor loss: -9.4876 Critic loss: +0.0091 World model loss: +0.5828 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5306 Episode_Reward/track_angular_velocity: +49.6659 Episode_Reward/upright: +25.2063 Episode_Reward/leg_posture: -2.4264 Episode_Reward/standstill: -5.5719 Episode_Reward/action_rate_l2: -0.5588 Episode_Reward/joint_torques_l2: -0.1409 Episode_Reward/dof_pos_limits: -0.0001 Episode_Reward/terminated: -0.0205 Episode_Reward/unnecessary_wheel_lift: -0.0618 Episode_Reward/rolling_leg_motion: -0.5255 Episode_Reward/obstacle_progress: +3.7070 Episode_Reward/wheel_body_intrusion: -0.2525 Episode_Reward/pair_body_intrusion: -0.2945 Curriculum/terrain_level_mean: +4.1614 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4365 Episode_Termination/fell_over: 0.0397 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0238 ============================================================================== ============================================================================== Learning iteration 39100/50000 Total steps: 160,153,600 Steps per second: 13,507 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1352.7 Actor loss: -9.5210 Critic loss: +0.0044 World model loss: +0.5563 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9308 Episode_Reward/track_angular_velocity: +51.0974 Episode_Reward/upright: +25.9184 Episode_Reward/leg_posture: -2.4971 Episode_Reward/standstill: -5.7778 Episode_Reward/action_rate_l2: -0.5895 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0878 Episode_Reward/rolling_leg_motion: -0.5772 Episode_Reward/obstacle_progress: +3.7850 Episode_Reward/wheel_body_intrusion: -0.2573 Episode_Reward/pair_body_intrusion: -0.2723 Curriculum/terrain_level_mean: +4.1626 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4906 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.4843 Episode_Termination/illegal_nonwheel_contact: 0.0126 ============================================================================== ============================================================================== Learning iteration 39150/50000 Total steps: 160,358,400 Steps per second: 13,510 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1424.6 Actor loss: -9.5457 Critic loss: +0.0040 World model loss: +0.5828 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.5707 Episode_Reward/track_angular_velocity: +53.8202 Episode_Reward/upright: +27.2939 Episode_Reward/leg_posture: -2.6752 Episode_Reward/standstill: -6.2883 Episode_Reward/action_rate_l2: -0.6070 Episode_Reward/joint_torques_l2: -0.1515 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0111 Episode_Reward/unnecessary_wheel_lift: -0.0823 Episode_Reward/rolling_leg_motion: -0.6086 Episode_Reward/obstacle_progress: +3.9073 Episode_Reward/wheel_body_intrusion: -0.2632 Episode_Reward/pair_body_intrusion: -0.2931 Curriculum/terrain_level_mean: +4.1636 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.5306 Episode_Termination/fell_over: 0.0204 Episode_Termination/course_exit: 0.4422 Episode_Termination/illegal_nonwheel_contact: 0.0136 ============================================================================== ============================================================================== Learning iteration 39200/50000 Total steps: 160,563,200 Steps per second: 13,512 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0877 Mean episode length: 1400.0 Actor loss: -9.5005 Critic loss: +0.0052 World model loss: +0.5752 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8218 Episode_Reward/track_angular_velocity: +52.8598 Episode_Reward/upright: +26.8223 Episode_Reward/leg_posture: -2.6575 Episode_Reward/standstill: -5.9594 Episode_Reward/action_rate_l2: -0.5828 Episode_Reward/joint_torques_l2: -0.1491 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0246 Episode_Reward/unnecessary_wheel_lift: -0.0950 Episode_Reward/rolling_leg_motion: -0.6029 Episode_Reward/obstacle_progress: +3.8590 Episode_Reward/wheel_body_intrusion: -0.2731 Episode_Reward/pair_body_intrusion: -0.2834 Curriculum/terrain_level_mean: +4.1646 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4713 Episode_Termination/fell_over: 0.0255 Episode_Termination/course_exit: 0.4713 Episode_Termination/illegal_nonwheel_contact: 0.0510 ============================================================================== ============================================================================== Learning iteration 39250/50000 Total steps: 160,768,000 Steps per second: 13,515 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1357.8 Actor loss: -9.4980 Critic loss: +0.0059 World model loss: +0.5838 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2806 Episode_Reward/track_angular_velocity: +51.3091 Episode_Reward/upright: +26.0208 Episode_Reward/leg_posture: -2.5429 Episode_Reward/standstill: -6.0713 Episode_Reward/action_rate_l2: -0.5686 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0851 Episode_Reward/rolling_leg_motion: -0.5728 Episode_Reward/obstacle_progress: +3.7656 Episode_Reward/wheel_body_intrusion: -0.2687 Episode_Reward/pair_body_intrusion: -0.2836 Curriculum/terrain_level_mean: +4.1660 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4667 Episode_Termination/fell_over: 0.0133 Episode_Termination/course_exit: 0.4800 Episode_Termination/illegal_nonwheel_contact: 0.0533 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[959] ============================================================================== Learning iteration 39300/50000 Total steps: 160,972,800 Steps per second: 13,517 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1367.7 Actor loss: -9.5032 Critic loss: +0.0061 World model loss: +0.5781 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4751 Episode_Reward/track_angular_velocity: +51.6036 Episode_Reward/upright: +26.2377 Episode_Reward/leg_posture: -2.5711 Episode_Reward/standstill: -6.1916 Episode_Reward/action_rate_l2: -0.5862 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0864 Episode_Reward/rolling_leg_motion: -0.5849 Episode_Reward/obstacle_progress: +3.7891 Episode_Reward/wheel_body_intrusion: -0.2722 Episode_Reward/pair_body_intrusion: -0.2590 Curriculum/terrain_level_mean: +4.1670 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4507 Episode_Termination/fell_over: 0.0352 Episode_Termination/course_exit: 0.5141 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== ============================================================================== Learning iteration 39350/50000 Total steps: 161,177,600 Steps per second: 13,519 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1329.7 Actor loss: -9.5148 Critic loss: +0.0049 World model loss: +0.5648 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1264 Episode_Reward/track_angular_velocity: +50.1419 Episode_Reward/upright: +25.5034 Episode_Reward/leg_posture: -2.5466 Episode_Reward/standstill: -6.2943 Episode_Reward/action_rate_l2: -0.5650 Episode_Reward/joint_torques_l2: -0.1425 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0390 Episode_Reward/unnecessary_wheel_lift: -0.0929 Episode_Reward/rolling_leg_motion: -0.5686 Episode_Reward/obstacle_progress: +3.7388 Episode_Reward/wheel_body_intrusion: -0.2614 Episode_Reward/pair_body_intrusion: -0.2293 Curriculum/terrain_level_mean: +4.1665 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.4539 Episode_Termination/fell_over: 0.0395 Episode_Termination/course_exit: 0.4474 Episode_Termination/illegal_nonwheel_contact: 0.0789 ============================================================================== ============================================================================== Learning iteration 39400/50000 Total steps: 161,382,400 Steps per second: 13,522 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1368.2 Actor loss: -9.5453 Critic loss: +0.0038 World model loss: +0.5597 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6348 Episode_Reward/track_angular_velocity: +51.7345 Episode_Reward/upright: +26.3059 Episode_Reward/leg_posture: -2.6127 Episode_Reward/standstill: -6.2613 Episode_Reward/action_rate_l2: -0.5783 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0899 Episode_Reward/rolling_leg_motion: -0.5851 Episode_Reward/obstacle_progress: +3.8533 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.2621 Curriculum/terrain_level_mean: +4.1694 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4904 Episode_Termination/fell_over: 0.0127 Episode_Termination/course_exit: 0.4777 Episode_Termination/illegal_nonwheel_contact: 0.0382 ============================================================================== ============================================================================== Learning iteration 39450/50000 Total steps: 161,587,200 Steps per second: 13,524 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1370.2 Actor loss: -9.4407 Critic loss: +0.0092 World model loss: +0.5620 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7080 Episode_Reward/track_angular_velocity: +51.7733 Episode_Reward/upright: +26.3472 Episode_Reward/leg_posture: -2.5737 Episode_Reward/standstill: -6.1117 Episode_Reward/action_rate_l2: -0.5763 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0820 Episode_Reward/rolling_leg_motion: -0.5823 Episode_Reward/obstacle_progress: +3.8699 Episode_Reward/wheel_body_intrusion: -0.2708 Episode_Reward/pair_body_intrusion: -0.2882 Curriculum/terrain_level_mean: +4.1719 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4303 Episode_Termination/fell_over: 0.0061 Episode_Termination/course_exit: 0.5515 Episode_Termination/illegal_nonwheel_contact: 0.0121 ============================================================================== ============================================================================== Learning iteration 39500/50000 Total steps: 161,792,000 Steps per second: 13,526 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1358.7 Actor loss: -9.5311 Critic loss: +0.0033 World model loss: +0.5662 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1852 Episode_Reward/track_angular_velocity: +51.2889 Episode_Reward/upright: +26.1010 Episode_Reward/leg_posture: -2.5568 Episode_Reward/standstill: -6.0981 Episode_Reward/action_rate_l2: -0.5800 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0018 Episode_Reward/terminated: -0.0106 Episode_Reward/unnecessary_wheel_lift: -0.0855 Episode_Reward/rolling_leg_motion: -0.5916 Episode_Reward/obstacle_progress: +3.8723 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.2738 Curriculum/terrain_level_mean: +4.1733 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4768 Episode_Termination/fell_over: 0.0132 Episode_Termination/course_exit: 0.4967 Episode_Termination/illegal_nonwheel_contact: 0.0132 ============================================================================== ============================================================================== Learning iteration 39550/50000 Total steps: 161,996,800 Steps per second: 13,528 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1376.6 Actor loss: -9.5039 Critic loss: +0.0065 World model loss: +0.5992 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8287 Episode_Reward/track_angular_velocity: +51.9707 Episode_Reward/upright: +26.4397 Episode_Reward/leg_posture: -2.5893 Episode_Reward/standstill: -6.3255 Episode_Reward/action_rate_l2: -0.5801 Episode_Reward/joint_torques_l2: -0.1457 Episode_Reward/dof_pos_limits: -0.0021 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0809 Episode_Reward/rolling_leg_motion: -0.5867 Episode_Reward/obstacle_progress: +3.9313 Episode_Reward/wheel_body_intrusion: -0.2553 Episode_Reward/pair_body_intrusion: -0.2825 Curriculum/terrain_level_mean: +4.1743 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4817 Episode_Termination/fell_over: 0.0183 Episode_Termination/course_exit: 0.4756 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 39600/50000 Total steps: 162,201,600 Steps per second: 13,531 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1356.6 Actor loss: -9.5367 Critic loss: +0.0069 World model loss: +0.5968 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1593 Episode_Reward/track_angular_velocity: +51.2341 Episode_Reward/upright: +26.0420 Episode_Reward/leg_posture: -2.5850 Episode_Reward/standstill: -6.1478 Episode_Reward/action_rate_l2: -0.5686 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0291 Episode_Reward/unnecessary_wheel_lift: -0.0641 Episode_Reward/rolling_leg_motion: -0.5603 Episode_Reward/obstacle_progress: +3.8720 Episode_Reward/wheel_body_intrusion: -0.2598 Episode_Reward/pair_body_intrusion: -0.3039 Curriculum/terrain_level_mean: +4.1743 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4370 Episode_Termination/fell_over: 0.0370 Episode_Termination/course_exit: 0.4815 Episode_Termination/illegal_nonwheel_contact: 0.0444 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[604] ============================================================================== Learning iteration 39650/50000 Total steps: 162,406,400 Steps per second: 13,533 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1338.0 Actor loss: -9.5506 Critic loss: +0.0061 World model loss: +0.5907 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4469 Episode_Reward/track_angular_velocity: +50.5246 Episode_Reward/upright: +25.6774 Episode_Reward/leg_posture: -2.5777 Episode_Reward/standstill: -6.3185 Episode_Reward/action_rate_l2: -0.5686 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0913 Episode_Reward/rolling_leg_motion: -0.5767 Episode_Reward/obstacle_progress: +3.8472 Episode_Reward/wheel_body_intrusion: -0.2762 Episode_Reward/pair_body_intrusion: -0.3028 Curriculum/terrain_level_mean: +4.1763 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4854 Episode_Termination/fell_over: 0.0234 Episode_Termination/course_exit: 0.4620 Episode_Termination/illegal_nonwheel_contact: 0.0409 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1639] ============================================================================== Learning iteration 39700/50000 Total steps: 162,611,200 Steps per second: 13,536 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1411.1 Actor loss: -9.5182 Critic loss: +0.0040 World model loss: +0.5982 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0171 Episode_Reward/track_angular_velocity: +53.2912 Episode_Reward/upright: +27.1069 Episode_Reward/leg_posture: -2.6765 Episode_Reward/standstill: -6.4004 Episode_Reward/action_rate_l2: -0.6066 Episode_Reward/joint_torques_l2: -0.1488 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.0940 Episode_Reward/rolling_leg_motion: -0.6253 Episode_Reward/obstacle_progress: +3.9388 Episode_Reward/wheel_body_intrusion: -0.2863 Episode_Reward/pair_body_intrusion: -0.2854 Curriculum/terrain_level_mean: +4.1785 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.5064 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.4615 Episode_Termination/illegal_nonwheel_contact: 0.0192 ============================================================================== ============================================================================== Learning iteration 39750/50000 Total steps: 162,816,000 Steps per second: 13,538 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1263.3 Actor loss: -9.5016 Critic loss: +0.0052 World model loss: +0.5958 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.8777 Episode_Reward/track_angular_velocity: +47.7164 Episode_Reward/upright: +24.2909 Episode_Reward/leg_posture: -2.3862 Episode_Reward/standstill: -5.8198 Episode_Reward/action_rate_l2: -0.5440 Episode_Reward/joint_torques_l2: -0.1330 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0173 Episode_Reward/unnecessary_wheel_lift: -0.0621 Episode_Reward/rolling_leg_motion: -0.5353 Episode_Reward/obstacle_progress: +3.5709 Episode_Reward/wheel_body_intrusion: -0.2565 Episode_Reward/pair_body_intrusion: -0.2385 Curriculum/terrain_level_mean: +4.1799 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.3803 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.5704 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2325] RECOVERY reset 1 invalid environments; first IDs=[1149] ============================================================================== Learning iteration 39800/50000 Total steps: 163,020,800 Steps per second: 13,540 Iteration time: 0.303s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0871 Mean episode length: 1331.3 Actor loss: -9.4858 Critic loss: +0.0047 World model loss: +0.5902 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1734 Episode_Reward/track_angular_velocity: +50.2593 Episode_Reward/upright: +25.5572 Episode_Reward/leg_posture: -2.5432 Episode_Reward/standstill: -6.1711 Episode_Reward/action_rate_l2: -0.5780 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0208 Episode_Reward/unnecessary_wheel_lift: -0.0853 Episode_Reward/rolling_leg_motion: -0.5828 Episode_Reward/obstacle_progress: +3.6882 Episode_Reward/wheel_body_intrusion: -0.2892 Episode_Reward/pair_body_intrusion: -0.2927 Curriculum/terrain_level_mean: +4.1804 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4014 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.5423 Episode_Termination/illegal_nonwheel_contact: 0.0423 ============================================================================== ============================================================================== Learning iteration 39850/50000 Total steps: 163,225,600 Steps per second: 13,543 Iteration time: 0.302s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1327.9 Actor loss: -9.5175 Critic loss: +0.0060 World model loss: +0.6099 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0849 Episode_Reward/track_angular_velocity: +50.0630 Episode_Reward/upright: +25.4911 Episode_Reward/leg_posture: -2.5505 Episode_Reward/standstill: -6.1153 Episode_Reward/action_rate_l2: -0.5818 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0233 Episode_Reward/unnecessary_wheel_lift: -0.0709 Episode_Reward/rolling_leg_motion: -0.5817 Episode_Reward/obstacle_progress: +3.6831 Episode_Reward/wheel_body_intrusion: -0.2743 Episode_Reward/pair_body_intrusion: -0.2833 Curriculum/terrain_level_mean: +4.1802 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1372 Episode_Termination/time_out: 0.4818 Episode_Termination/fell_over: 0.0146 Episode_Termination/course_exit: 0.4526 Episode_Termination/illegal_nonwheel_contact: 0.0511 ============================================================================== ============================================================================== Learning iteration 39900/50000 Total steps: 163,430,400 Steps per second: 13,545 Iteration time: 0.302s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1310.4 Actor loss: -9.5377 Critic loss: +0.0052 World model loss: +0.5701 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4007 Episode_Reward/track_angular_velocity: +49.4922 Episode_Reward/upright: +25.1490 Episode_Reward/leg_posture: -2.4953 Episode_Reward/standstill: -5.6753 Episode_Reward/action_rate_l2: -0.5714 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0154 Episode_Reward/unnecessary_wheel_lift: -0.0726 Episode_Reward/rolling_leg_motion: -0.5556 Episode_Reward/obstacle_progress: +3.6855 Episode_Reward/wheel_body_intrusion: -0.2587 Episode_Reward/pair_body_intrusion: -0.2889 Curriculum/terrain_level_mean: +4.1809 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4419 Episode_Termination/fell_over: 0.0078 Episode_Termination/course_exit: 0.5116 Episode_Termination/illegal_nonwheel_contact: 0.0465 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[945] ============================================================================== Learning iteration 39950/50000 Total steps: 163,635,200 Steps per second: 13,547 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1409.7 Actor loss: -9.4939 Critic loss: +0.0081 World model loss: +0.5776 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9639 Episode_Reward/track_angular_velocity: +53.2454 Episode_Reward/upright: +27.0561 Episode_Reward/leg_posture: -2.6394 Episode_Reward/standstill: -6.0792 Episode_Reward/action_rate_l2: -0.6045 Episode_Reward/joint_torques_l2: -0.1506 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0220 Episode_Reward/unnecessary_wheel_lift: -0.0765 Episode_Reward/rolling_leg_motion: -0.5910 Episode_Reward/obstacle_progress: +4.0660 Episode_Reward/wheel_body_intrusion: -0.2725 Episode_Reward/pair_body_intrusion: -0.2842 Curriculum/terrain_level_mean: +4.1814 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5113 Episode_Termination/fell_over: 0.0376 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0226 ============================================================================== ============================================================================== Learning iteration 40000/50000 Total steps: 163,840,000 Steps per second: 13,550 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1386.3 Actor loss: -9.4951 Critic loss: +0.0084 World model loss: +0.5769 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0404 Episode_Reward/track_angular_velocity: +52.3856 Episode_Reward/upright: +26.6176 Episode_Reward/leg_posture: -2.5635 Episode_Reward/standstill: -6.2172 Episode_Reward/action_rate_l2: -0.5976 Episode_Reward/joint_torques_l2: -0.1485 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0244 Episode_Reward/unnecessary_wheel_lift: -0.0980 Episode_Reward/rolling_leg_motion: -0.5855 Episode_Reward/obstacle_progress: +4.0189 Episode_Reward/wheel_body_intrusion: -0.2522 Episode_Reward/pair_body_intrusion: -0.2766 Curriculum/terrain_level_mean: +4.1831 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4885 Episode_Termination/fell_over: 0.0172 Episode_Termination/course_exit: 0.4598 Episode_Termination/illegal_nonwheel_contact: 0.0460 ============================================================================== ============================================================================== Learning iteration 40050/50000 Total steps: 164,044,800 Steps per second: 13,552 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1369.8 Actor loss: -9.5112 Critic loss: +0.0046 World model loss: +0.5735 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4591 Episode_Reward/track_angular_velocity: +51.7530 Episode_Reward/upright: +26.3161 Episode_Reward/leg_posture: -2.5402 Episode_Reward/standstill: -6.1224 Episode_Reward/action_rate_l2: -0.5887 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.0880 Episode_Reward/rolling_leg_motion: -0.5937 Episode_Reward/obstacle_progress: +3.8494 Episode_Reward/wheel_body_intrusion: -0.2591 Episode_Reward/pair_body_intrusion: -0.2708 Curriculum/terrain_level_mean: +4.1838 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4914 Episode_Termination/fell_over: 0.0286 Episode_Termination/course_exit: 0.4571 Episode_Termination/illegal_nonwheel_contact: 0.0400 ============================================================================== ============================================================================== Learning iteration 40100/50000 Total steps: 164,249,600 Steps per second: 13,554 Iteration time: 0.302s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0878 Mean episode length: 1375.8 Actor loss: -9.4904 Critic loss: +0.0051 World model loss: +0.5764 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9136 Episode_Reward/track_angular_velocity: +52.0212 Episode_Reward/upright: +26.4367 Episode_Reward/leg_posture: -2.5581 Episode_Reward/standstill: -6.0466 Episode_Reward/action_rate_l2: -0.5912 Episode_Reward/joint_torques_l2: -0.1471 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0269 Episode_Reward/unnecessary_wheel_lift: -0.0746 Episode_Reward/rolling_leg_motion: -0.5885 Episode_Reward/obstacle_progress: +3.9404 Episode_Reward/wheel_body_intrusion: -0.2480 Episode_Reward/pair_body_intrusion: -0.2624 Curriculum/terrain_level_mean: +4.1841 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4895 Episode_Termination/fell_over: 0.0350 Episode_Termination/course_exit: 0.4406 Episode_Termination/illegal_nonwheel_contact: 0.0350 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1274] ============================================================================== Learning iteration 40150/50000 Total steps: 164,454,400 Steps per second: 13,557 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0877 Mean episode length: 1376.8 Actor loss: -9.5384 Critic loss: +0.0069 World model loss: +0.6298 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9542 Episode_Reward/track_angular_velocity: +52.0873 Episode_Reward/upright: +26.4100 Episode_Reward/leg_posture: -2.5959 Episode_Reward/standstill: -6.0556 Episode_Reward/action_rate_l2: -0.5857 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0054 Episode_Reward/unnecessary_wheel_lift: -0.0966 Episode_Reward/rolling_leg_motion: -0.6004 Episode_Reward/obstacle_progress: +3.9400 Episode_Reward/wheel_body_intrusion: -0.2708 Episode_Reward/pair_body_intrusion: -0.2885 Curriculum/terrain_level_mean: +4.1875 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4862 Episode_Termination/fell_over: 0.0055 Episode_Termination/course_exit: 0.5028 Episode_Termination/illegal_nonwheel_contact: 0.0055 ============================================================================== ============================================================================== Learning iteration 40200/50000 Total steps: 164,659,200 Steps per second: 13,559 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1334.4 Actor loss: -9.5170 Critic loss: +0.0053 World model loss: +0.5696 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3501 Episode_Reward/track_angular_velocity: +50.4151 Episode_Reward/upright: +25.5872 Episode_Reward/leg_posture: -2.5265 Episode_Reward/standstill: -6.0763 Episode_Reward/action_rate_l2: -0.5683 Episode_Reward/joint_torques_l2: -0.1430 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0940 Episode_Reward/rolling_leg_motion: -0.5763 Episode_Reward/obstacle_progress: +3.7783 Episode_Reward/wheel_body_intrusion: -0.2797 Episode_Reward/pair_body_intrusion: -0.3276 Curriculum/terrain_level_mean: +4.1882 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4379 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.5033 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 40250/50000 Total steps: 164,864,000 Steps per second: 13,561 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1356.1 Actor loss: -9.5173 Critic loss: +0.0048 World model loss: +0.5849 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1323 Episode_Reward/track_angular_velocity: +51.3190 Episode_Reward/upright: +26.0123 Episode_Reward/leg_posture: -2.5076 Episode_Reward/standstill: -6.1375 Episode_Reward/action_rate_l2: -0.5826 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0291 Episode_Reward/unnecessary_wheel_lift: -0.0904 Episode_Reward/rolling_leg_motion: -0.5918 Episode_Reward/obstacle_progress: +3.8433 Episode_Reward/wheel_body_intrusion: -0.2793 Episode_Reward/pair_body_intrusion: -0.3360 Curriculum/terrain_level_mean: +4.1892 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4970 Episode_Termination/fell_over: 0.0355 Episode_Termination/course_exit: 0.4379 Episode_Termination/illegal_nonwheel_contact: 0.0473 ============================================================================== ============================================================================== Learning iteration 40300/50000 Total steps: 165,068,800 Steps per second: 13,564 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1321.4 Actor loss: -9.5109 Critic loss: +0.0035 World model loss: +0.5638 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9050 Episode_Reward/track_angular_velocity: +49.9488 Episode_Reward/upright: +25.3191 Episode_Reward/leg_posture: -2.4770 Episode_Reward/standstill: -5.9305 Episode_Reward/action_rate_l2: -0.5574 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0867 Episode_Reward/rolling_leg_motion: -0.5554 Episode_Reward/obstacle_progress: +3.7028 Episode_Reward/wheel_body_intrusion: -0.2710 Episode_Reward/pair_body_intrusion: -0.2474 Curriculum/terrain_level_mean: +4.1897 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4317 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.5108 Episode_Termination/illegal_nonwheel_contact: 0.0360 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2034] ============================================================================== Learning iteration 40350/50000 Total steps: 165,273,600 Steps per second: 13,566 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0873 Mean episode length: 1373.5 Actor loss: -9.5408 Critic loss: +0.0041 World model loss: +0.5522 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7280 Episode_Reward/track_angular_velocity: +51.9026 Episode_Reward/upright: +26.3082 Episode_Reward/leg_posture: -2.5885 Episode_Reward/standstill: -6.2332 Episode_Reward/action_rate_l2: -0.5868 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.0918 Episode_Reward/rolling_leg_motion: -0.5927 Episode_Reward/obstacle_progress: +3.7446 Episode_Reward/wheel_body_intrusion: -0.2685 Episode_Reward/pair_body_intrusion: -0.2641 Curriculum/terrain_level_mean: +4.1909 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4323 Episode_Termination/fell_over: 0.0323 Episode_Termination/course_exit: 0.5097 Episode_Termination/illegal_nonwheel_contact: 0.0387 ============================================================================== ============================================================================== Learning iteration 40400/50000 Total steps: 165,478,400 Steps per second: 13,568 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1358.0 Actor loss: -9.5107 Critic loss: +0.0053 World model loss: +0.5693 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2827 Episode_Reward/track_angular_velocity: +51.3329 Episode_Reward/upright: +25.9727 Episode_Reward/leg_posture: -2.5512 Episode_Reward/standstill: -5.9682 Episode_Reward/action_rate_l2: -0.5746 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0690 Episode_Reward/rolling_leg_motion: -0.5784 Episode_Reward/obstacle_progress: +3.7569 Episode_Reward/wheel_body_intrusion: -0.2643 Episode_Reward/pair_body_intrusion: -0.2802 Curriculum/terrain_level_mean: +4.1909 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4414 Episode_Termination/fell_over: 0.0276 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 40450/50000 Total steps: 165,683,200 Steps per second: 13,570 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1323.8 Actor loss: -9.5245 Critic loss: +0.0044 World model loss: +0.5716 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8602 Episode_Reward/track_angular_velocity: +50.0028 Episode_Reward/upright: +25.3130 Episode_Reward/leg_posture: -2.4822 Episode_Reward/standstill: -6.2612 Episode_Reward/action_rate_l2: -0.5675 Episode_Reward/joint_torques_l2: -0.1418 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0852 Episode_Reward/rolling_leg_motion: -0.5744 Episode_Reward/obstacle_progress: +3.6051 Episode_Reward/wheel_body_intrusion: -0.2593 Episode_Reward/pair_body_intrusion: -0.2572 Curriculum/terrain_level_mean: +4.1931 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4218 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.5374 Episode_Termination/illegal_nonwheel_contact: 0.0340 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1598] RECOVERY reset 1 invalid environments; first IDs=[2193] ============================================================================== Learning iteration 40500/50000 Total steps: 165,888,000 Steps per second: 13,572 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1358.1 Actor loss: -9.5082 Critic loss: +0.0046 World model loss: +0.5705 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0348 Episode_Reward/track_angular_velocity: +51.2939 Episode_Reward/upright: +25.9517 Episode_Reward/leg_posture: -2.5894 Episode_Reward/standstill: -6.3810 Episode_Reward/action_rate_l2: -0.5779 Episode_Reward/joint_torques_l2: -0.1467 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0292 Episode_Reward/unnecessary_wheel_lift: -0.0973 Episode_Reward/rolling_leg_motion: -0.5812 Episode_Reward/obstacle_progress: +3.7598 Episode_Reward/wheel_body_intrusion: -0.2701 Episode_Reward/pair_body_intrusion: -0.2966 Curriculum/terrain_level_mean: +4.1934 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4221 Episode_Termination/fell_over: 0.0455 Episode_Termination/course_exit: 0.5065 Episode_Termination/illegal_nonwheel_contact: 0.0390 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_040500.mp4 ============================================================================== Learning iteration 40550/50000 Total steps: 166,092,800 Steps per second: 13,568 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0873 Mean episode length: 1364.3 Actor loss: -9.5033 Critic loss: +0.0054 World model loss: +0.5980 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3475 Episode_Reward/track_angular_velocity: +51.5452 Episode_Reward/upright: +26.1035 Episode_Reward/leg_posture: -2.5639 Episode_Reward/standstill: -6.2445 Episode_Reward/action_rate_l2: -0.5827 Episode_Reward/joint_torques_l2: -0.1465 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0767 Episode_Reward/rolling_leg_motion: -0.5607 Episode_Reward/obstacle_progress: +3.8043 Episode_Reward/wheel_body_intrusion: -0.2600 Episode_Reward/pair_body_intrusion: -0.3089 Curriculum/terrain_level_mean: +4.1941 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4853 Episode_Termination/fell_over: 0.0294 Episode_Termination/course_exit: 0.4706 Episode_Termination/illegal_nonwheel_contact: 0.0147 ============================================================================== ============================================================================== Learning iteration 40600/50000 Total steps: 166,297,600 Steps per second: 13,570 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1319.9 Actor loss: -9.4965 Critic loss: +0.0040 World model loss: +0.5910 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6926 Episode_Reward/track_angular_velocity: +49.8798 Episode_Reward/upright: +25.2533 Episode_Reward/leg_posture: -2.4697 Episode_Reward/standstill: -5.8938 Episode_Reward/action_rate_l2: -0.5627 Episode_Reward/joint_torques_l2: -0.1433 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0284 Episode_Reward/unnecessary_wheel_lift: -0.0775 Episode_Reward/rolling_leg_motion: -0.5503 Episode_Reward/obstacle_progress: +3.6810 Episode_Reward/wheel_body_intrusion: -0.2678 Episode_Reward/pair_body_intrusion: -0.2863 Curriculum/terrain_level_mean: +4.1946 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4452 Episode_Termination/fell_over: 0.0194 Episode_Termination/course_exit: 0.4839 Episode_Termination/illegal_nonwheel_contact: 0.0516 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1902] ============================================================================== Learning iteration 40650/50000 Total steps: 166,502,400 Steps per second: 13,573 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1376.6 Actor loss: -9.5207 Critic loss: +0.0045 World model loss: +0.5841 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7818 Episode_Reward/track_angular_velocity: +51.9337 Episode_Reward/upright: +26.3524 Episode_Reward/leg_posture: -2.6281 Episode_Reward/standstill: -6.3200 Episode_Reward/action_rate_l2: -0.5858 Episode_Reward/joint_torques_l2: -0.1500 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.1025 Episode_Reward/rolling_leg_motion: -0.5822 Episode_Reward/obstacle_progress: +3.8465 Episode_Reward/wheel_body_intrusion: -0.2714 Episode_Reward/pair_body_intrusion: -0.3073 Curriculum/terrain_level_mean: +4.1970 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.5147 Episode_Termination/fell_over: 0.0074 Episode_Termination/course_exit: 0.4412 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 40700/50000 Total steps: 166,707,200 Steps per second: 13,575 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1263.1 Actor loss: -9.5101 Critic loss: +0.0059 World model loss: +0.5769 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.7567 Episode_Reward/track_angular_velocity: +47.6787 Episode_Reward/upright: +24.1637 Episode_Reward/leg_posture: -2.3814 Episode_Reward/standstill: -5.7310 Episode_Reward/action_rate_l2: -0.5460 Episode_Reward/joint_torques_l2: -0.1373 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0247 Episode_Reward/unnecessary_wheel_lift: -0.0868 Episode_Reward/rolling_leg_motion: -0.5289 Episode_Reward/obstacle_progress: +3.5295 Episode_Reward/wheel_body_intrusion: -0.2445 Episode_Reward/pair_body_intrusion: -0.2856 Curriculum/terrain_level_mean: +4.1978 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.3724 Episode_Termination/fell_over: 0.0552 Episode_Termination/course_exit: 0.5586 Episode_Termination/illegal_nonwheel_contact: 0.0207 ============================================================================== ============================================================================== Learning iteration 40750/50000 Total steps: 166,912,000 Steps per second: 13,577 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0868 Mean episode length: 1353.5 Actor loss: -9.4917 Critic loss: +0.0056 World model loss: +0.5854 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8815 Episode_Reward/track_angular_velocity: +51.1224 Episode_Reward/upright: +25.9305 Episode_Reward/leg_posture: -2.5655 Episode_Reward/standstill: -6.5737 Episode_Reward/action_rate_l2: -0.5831 Episode_Reward/joint_torques_l2: -0.1457 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0317 Episode_Reward/unnecessary_wheel_lift: -0.1032 Episode_Reward/rolling_leg_motion: -0.5710 Episode_Reward/obstacle_progress: +3.7029 Episode_Reward/wheel_body_intrusion: -0.2671 Episode_Reward/pair_body_intrusion: -0.2861 Curriculum/terrain_level_mean: +4.1987 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4671 Episode_Termination/fell_over: 0.0359 Episode_Termination/course_exit: 0.4611 Episode_Termination/illegal_nonwheel_contact: 0.0599 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1653] ============================================================================== Learning iteration 40800/50000 Total steps: 167,116,800 Steps per second: 13,579 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1321.8 Actor loss: -9.4924 Critic loss: +0.0059 World model loss: +0.5598 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8680 Episode_Reward/track_angular_velocity: +49.9450 Episode_Reward/upright: +25.3599 Episode_Reward/leg_posture: -2.5074 Episode_Reward/standstill: -5.7819 Episode_Reward/action_rate_l2: -0.5514 Episode_Reward/joint_torques_l2: -0.1414 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0129 Episode_Reward/unnecessary_wheel_lift: -0.0646 Episode_Reward/rolling_leg_motion: -0.5504 Episode_Reward/obstacle_progress: +3.6378 Episode_Reward/wheel_body_intrusion: -0.2761 Episode_Reward/pair_body_intrusion: -0.2469 Curriculum/terrain_level_mean: +4.1982 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4775 Episode_Termination/fell_over: 0.0090 Episode_Termination/course_exit: 0.4775 Episode_Termination/illegal_nonwheel_contact: 0.0450 ============================================================================== ============================================================================== Learning iteration 40850/50000 Total steps: 167,321,600 Steps per second: 13,582 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0871 Mean episode length: 1344.8 Actor loss: -9.5320 Critic loss: +0.0036 World model loss: +0.5700 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5282 Episode_Reward/track_angular_velocity: +50.7710 Episode_Reward/upright: +25.7732 Episode_Reward/leg_posture: -2.5577 Episode_Reward/standstill: -6.1757 Episode_Reward/action_rate_l2: -0.5739 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0263 Episode_Reward/unnecessary_wheel_lift: -0.0881 Episode_Reward/rolling_leg_motion: -0.5759 Episode_Reward/obstacle_progress: +3.7432 Episode_Reward/wheel_body_intrusion: -0.2841 Episode_Reward/pair_body_intrusion: -0.2776 Curriculum/terrain_level_mean: +4.1997 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4601 Episode_Termination/fell_over: 0.0184 Episode_Termination/course_exit: 0.4785 Episode_Termination/illegal_nonwheel_contact: 0.0491 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[3562] ============================================================================== Learning iteration 40900/50000 Total steps: 167,526,400 Steps per second: 13,584 Iteration time: 0.302s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0873 Mean episode length: 1384.6 Actor loss: -9.5143 Critic loss: +0.0069 World model loss: +0.5512 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0059 Episode_Reward/track_angular_velocity: +52.2522 Episode_Reward/upright: +26.5421 Episode_Reward/leg_posture: -2.6669 Episode_Reward/standstill: -6.0371 Episode_Reward/action_rate_l2: -0.5905 Episode_Reward/joint_torques_l2: -0.1498 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0216 Episode_Reward/unnecessary_wheel_lift: -0.0887 Episode_Reward/rolling_leg_motion: -0.5750 Episode_Reward/obstacle_progress: +3.8621 Episode_Reward/wheel_body_intrusion: -0.3039 Episode_Reward/pair_body_intrusion: -0.3310 Curriculum/terrain_level_mean: +4.2002 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4823 Episode_Termination/fell_over: 0.0426 Episode_Termination/course_exit: 0.4610 Episode_Termination/illegal_nonwheel_contact: 0.0284 ============================================================================== ============================================================================== Learning iteration 40950/50000 Total steps: 167,731,200 Steps per second: 13,586 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1429.0 Actor loss: -9.5040 Critic loss: +0.0048 World model loss: +0.5751 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7174 Episode_Reward/track_angular_velocity: +53.9905 Episode_Reward/upright: +27.3772 Episode_Reward/leg_posture: -2.7239 Episode_Reward/standstill: -6.2677 Episode_Reward/action_rate_l2: -0.6105 Episode_Reward/joint_torques_l2: -0.1530 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0223 Episode_Reward/unnecessary_wheel_lift: -0.0859 Episode_Reward/rolling_leg_motion: -0.6177 Episode_Reward/obstacle_progress: +4.0260 Episode_Reward/wheel_body_intrusion: -0.2863 Episode_Reward/pair_body_intrusion: -0.3068 Curriculum/terrain_level_mean: +4.2012 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5309 Episode_Termination/fell_over: 0.0370 Episode_Termination/course_exit: 0.4198 Episode_Termination/illegal_nonwheel_contact: 0.0247 ============================================================================== ============================================================================== Learning iteration 41000/50000 Total steps: 167,936,000 Steps per second: 13,588 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0874 Mean episode length: 1423.3 Actor loss: -9.5314 Critic loss: +0.0046 World model loss: +0.5804 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.5307 Episode_Reward/track_angular_velocity: +53.7867 Episode_Reward/upright: +27.3012 Episode_Reward/leg_posture: -2.7062 Episode_Reward/standstill: -6.4071 Episode_Reward/action_rate_l2: -0.6055 Episode_Reward/joint_torques_l2: -0.1507 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0334 Episode_Reward/unnecessary_wheel_lift: -0.0962 Episode_Reward/rolling_leg_motion: -0.6272 Episode_Reward/obstacle_progress: +3.9742 Episode_Reward/wheel_body_intrusion: -0.2695 Episode_Reward/pair_body_intrusion: -0.3081 Curriculum/terrain_level_mean: +4.1995 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.5101 Episode_Termination/fell_over: 0.0604 Episode_Termination/course_exit: 0.4094 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 41050/50000 Total steps: 168,140,800 Steps per second: 13,590 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1291.1 Actor loss: -9.5351 Critic loss: +0.0029 World model loss: +0.5462 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8293 Episode_Reward/track_angular_velocity: +48.8176 Episode_Reward/upright: +24.7563 Episode_Reward/leg_posture: -2.4670 Episode_Reward/standstill: -5.6363 Episode_Reward/action_rate_l2: -0.5453 Episode_Reward/joint_torques_l2: -0.1381 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0177 Episode_Reward/unnecessary_wheel_lift: -0.0853 Episode_Reward/rolling_leg_motion: -0.5662 Episode_Reward/obstacle_progress: +3.6109 Episode_Reward/wheel_body_intrusion: -0.2585 Episode_Reward/pair_body_intrusion: -0.2541 Curriculum/terrain_level_mean: +4.2012 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4156 Episode_Termination/fell_over: 0.0260 Episode_Termination/course_exit: 0.5390 Episode_Termination/illegal_nonwheel_contact: 0.0325 ============================================================================== ============================================================================== Learning iteration 41100/50000 Total steps: 168,345,600 Steps per second: 13,592 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1376.7 Actor loss: -9.5108 Critic loss: +0.0073 World model loss: +0.5591 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9279 Episode_Reward/track_angular_velocity: +52.0831 Episode_Reward/upright: +26.3885 Episode_Reward/leg_posture: -2.5942 Episode_Reward/standstill: -6.0365 Episode_Reward/action_rate_l2: -0.5861 Episode_Reward/joint_torques_l2: -0.1458 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0323 Episode_Reward/unnecessary_wheel_lift: -0.0907 Episode_Reward/rolling_leg_motion: -0.5819 Episode_Reward/obstacle_progress: +3.9760 Episode_Reward/wheel_body_intrusion: -0.2572 Episode_Reward/pair_body_intrusion: -0.2467 Curriculum/terrain_level_mean: +4.1995 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1333 Episode_Termination/time_out: 0.5098 Episode_Termination/fell_over: 0.0458 Episode_Termination/course_exit: 0.4118 Episode_Termination/illegal_nonwheel_contact: 0.0588 ============================================================================== ============================================================================== Learning iteration 41150/50000 Total steps: 168,550,400 Steps per second: 13,595 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1323.4 Actor loss: -9.5135 Critic loss: +0.0048 World model loss: +0.5685 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8799 Episode_Reward/track_angular_velocity: +49.9982 Episode_Reward/upright: +25.3614 Episode_Reward/leg_posture: -2.5025 Episode_Reward/standstill: -5.7957 Episode_Reward/action_rate_l2: -0.5723 Episode_Reward/joint_torques_l2: -0.1403 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0129 Episode_Reward/unnecessary_wheel_lift: -0.0828 Episode_Reward/rolling_leg_motion: -0.5726 Episode_Reward/obstacle_progress: +3.7892 Episode_Reward/wheel_body_intrusion: -0.2626 Episode_Reward/pair_body_intrusion: -0.2835 Curriculum/terrain_level_mean: +4.2017 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4662 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0068 ============================================================================== ============================================================================== Learning iteration 41200/50000 Total steps: 168,755,200 Steps per second: 13,597 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0877 Mean episode length: 1333.3 Actor loss: -9.5373 Critic loss: +0.0035 World model loss: +0.5934 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2212 Episode_Reward/track_angular_velocity: +50.3797 Episode_Reward/upright: +25.5404 Episode_Reward/leg_posture: -2.4851 Episode_Reward/standstill: -5.6795 Episode_Reward/action_rate_l2: -0.5771 Episode_Reward/joint_torques_l2: -0.1409 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0417 Episode_Reward/unnecessary_wheel_lift: -0.1024 Episode_Reward/rolling_leg_motion: -0.5742 Episode_Reward/obstacle_progress: +3.8688 Episode_Reward/wheel_body_intrusion: -0.2551 Episode_Reward/pair_body_intrusion: -0.2458 Curriculum/terrain_level_mean: +4.1992 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4293 Episode_Termination/fell_over: 0.0598 Episode_Termination/course_exit: 0.4837 Episode_Termination/illegal_nonwheel_contact: 0.0598 ============================================================================== ============================================================================== Learning iteration 41250/50000 Total steps: 168,960,000 Steps per second: 13,599 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0882 Mean episode length: 1374.7 Actor loss: -9.5292 Critic loss: +0.0043 World model loss: +0.5595 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9935 Episode_Reward/track_angular_velocity: +52.0652 Episode_Reward/upright: +26.4142 Episode_Reward/leg_posture: -2.5543 Episode_Reward/standstill: -5.7792 Episode_Reward/action_rate_l2: -0.5802 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0746 Episode_Reward/rolling_leg_motion: -0.5810 Episode_Reward/obstacle_progress: +3.9449 Episode_Reward/wheel_body_intrusion: -0.2660 Episode_Reward/pair_body_intrusion: -0.2466 Curriculum/terrain_level_mean: +4.2000 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1320 Episode_Termination/time_out: 0.5164 Episode_Termination/fell_over: 0.0082 Episode_Termination/course_exit: 0.4426 Episode_Termination/illegal_nonwheel_contact: 0.0410 ============================================================================== ============================================================================== Learning iteration 41300/50000 Total steps: 169,164,800 Steps per second: 13,601 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1384.0 Actor loss: -9.5205 Critic loss: +0.0052 World model loss: +0.6114 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1060 Episode_Reward/track_angular_velocity: +52.3245 Episode_Reward/upright: +26.5469 Episode_Reward/leg_posture: -2.5692 Episode_Reward/standstill: -5.9087 Episode_Reward/action_rate_l2: -0.5850 Episode_Reward/joint_torques_l2: -0.1468 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0216 Episode_Reward/unnecessary_wheel_lift: -0.0794 Episode_Reward/rolling_leg_motion: -0.5885 Episode_Reward/obstacle_progress: +3.9593 Episode_Reward/wheel_body_intrusion: -0.2756 Episode_Reward/pair_body_intrusion: -0.2524 Curriculum/terrain_level_mean: +4.2000 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4870 Episode_Termination/fell_over: 0.0195 Episode_Termination/course_exit: 0.4610 Episode_Termination/illegal_nonwheel_contact: 0.0455 ============================================================================== ============================================================================== Learning iteration 41350/50000 Total steps: 169,369,600 Steps per second: 13,603 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1360.1 Actor loss: -9.5188 Critic loss: +0.0077 World model loss: +0.5621 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2442 Episode_Reward/track_angular_velocity: +51.3780 Episode_Reward/upright: +26.0623 Episode_Reward/leg_posture: -2.5869 Episode_Reward/standstill: -5.9159 Episode_Reward/action_rate_l2: -0.5720 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0186 Episode_Reward/unnecessary_wheel_lift: -0.0730 Episode_Reward/rolling_leg_motion: -0.5802 Episode_Reward/obstacle_progress: +3.8489 Episode_Reward/wheel_body_intrusion: -0.2964 Episode_Reward/pair_body_intrusion: -0.2932 Curriculum/terrain_level_mean: +4.2012 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4857 Episode_Termination/fell_over: 0.0214 Episode_Termination/course_exit: 0.4643 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== ============================================================================== Learning iteration 41400/50000 Total steps: 169,574,400 Steps per second: 13,606 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1370.6 Actor loss: -9.5208 Critic loss: +0.0073 World model loss: +0.5474 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7461 Episode_Reward/track_angular_velocity: +51.8089 Episode_Reward/upright: +26.2513 Episode_Reward/leg_posture: -2.5913 Episode_Reward/standstill: -6.2809 Episode_Reward/action_rate_l2: -0.5694 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0849 Episode_Reward/rolling_leg_motion: -0.5770 Episode_Reward/obstacle_progress: +3.7587 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.2731 Curriculum/terrain_level_mean: +4.2036 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4524 Episode_Termination/fell_over: 0.0298 Episode_Termination/course_exit: 0.5060 Episode_Termination/illegal_nonwheel_contact: 0.0179 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1964] RECOVERY reset 1 invalid environments; first IDs=[1085] ============================================================================== Learning iteration 41450/50000 Total steps: 169,779,200 Steps per second: 13,608 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1362.7 Actor loss: -9.5261 Critic loss: +0.0052 World model loss: +0.5651 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4364 Episode_Reward/track_angular_velocity: +51.4822 Episode_Reward/upright: +26.1149 Episode_Reward/leg_posture: -2.5675 Episode_Reward/standstill: -6.0369 Episode_Reward/action_rate_l2: -0.5617 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0160 Episode_Reward/unnecessary_wheel_lift: -0.0751 Episode_Reward/rolling_leg_motion: -0.5793 Episode_Reward/obstacle_progress: +3.7196 Episode_Reward/wheel_body_intrusion: -0.2656 Episode_Reward/pair_body_intrusion: -0.3045 Curriculum/terrain_level_mean: +4.2043 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4621 Episode_Termination/fell_over: 0.0138 Episode_Termination/course_exit: 0.4966 Episode_Termination/illegal_nonwheel_contact: 0.0345 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2459] ============================================================================== Learning iteration 41500/50000 Total steps: 169,984,000 Steps per second: 13,610 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0870 Mean episode length: 1393.9 Actor loss: -9.5274 Critic loss: +0.0042 World model loss: +0.5600 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.4519 Episode_Reward/track_angular_velocity: +52.5961 Episode_Reward/upright: +26.6846 Episode_Reward/leg_posture: -2.6385 Episode_Reward/standstill: -6.6633 Episode_Reward/action_rate_l2: -0.5824 Episode_Reward/joint_torques_l2: -0.1472 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0245 Episode_Reward/unnecessary_wheel_lift: -0.0695 Episode_Reward/rolling_leg_motion: -0.5734 Episode_Reward/obstacle_progress: +3.7933 Episode_Reward/wheel_body_intrusion: -0.2764 Episode_Reward/pair_body_intrusion: -0.3184 Curriculum/terrain_level_mean: +4.2031 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4820 Episode_Termination/fell_over: 0.0432 Episode_Termination/course_exit: 0.4532 Episode_Termination/illegal_nonwheel_contact: 0.0504 ============================================================================== ============================================================================== Learning iteration 41550/50000 Total steps: 170,188,800 Steps per second: 13,612 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0869 Mean episode length: 1408.7 Actor loss: -9.5126 Critic loss: +0.0040 World model loss: +0.6071 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8279 Episode_Reward/track_angular_velocity: +53.0880 Episode_Reward/upright: +26.9799 Episode_Reward/leg_posture: -2.7058 Episode_Reward/standstill: -6.6671 Episode_Reward/action_rate_l2: -0.5920 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0055 Episode_Reward/unnecessary_wheel_lift: -0.0696 Episode_Reward/rolling_leg_motion: -0.5822 Episode_Reward/obstacle_progress: +3.8134 Episode_Reward/wheel_body_intrusion: -0.2687 Episode_Reward/pair_body_intrusion: -0.3180 Curriculum/terrain_level_mean: +4.2056 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1377 Episode_Termination/time_out: 0.4964 Episode_Termination/fell_over: 0.0072 Episode_Termination/course_exit: 0.4892 Episode_Termination/illegal_nonwheel_contact: 0.0072 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2299] ============================================================================== Learning iteration 41600/50000 Total steps: 170,393,600 Steps per second: 13,614 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0870 Mean episode length: 1296.7 Actor loss: -9.5015 Critic loss: +0.0038 World model loss: +0.5589 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.8729 Episode_Reward/track_angular_velocity: +48.8975 Episode_Reward/upright: +24.8206 Episode_Reward/leg_posture: -2.5246 Episode_Reward/standstill: -5.9304 Episode_Reward/action_rate_l2: -0.5530 Episode_Reward/joint_torques_l2: -0.1383 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0177 Episode_Reward/unnecessary_wheel_lift: -0.0856 Episode_Reward/rolling_leg_motion: -0.5573 Episode_Reward/obstacle_progress: +3.5380 Episode_Reward/wheel_body_intrusion: -0.2651 Episode_Reward/pair_body_intrusion: -0.2717 Curriculum/terrain_level_mean: +4.2073 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.4178 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.5342 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 41650/50000 Total steps: 170,598,400 Steps per second: 13,616 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1323.7 Actor loss: -9.5108 Critic loss: +0.0041 World model loss: +0.5628 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7907 Episode_Reward/track_angular_velocity: +49.9332 Episode_Reward/upright: +25.3318 Episode_Reward/leg_posture: -2.5543 Episode_Reward/standstill: -6.4918 Episode_Reward/action_rate_l2: -0.5501 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.0711 Episode_Reward/rolling_leg_motion: -0.5613 Episode_Reward/obstacle_progress: +3.6368 Episode_Reward/wheel_body_intrusion: -0.2662 Episode_Reward/pair_body_intrusion: -0.2845 Curriculum/terrain_level_mean: +4.2092 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0948 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4286 Episode_Termination/fell_over: 0.0357 Episode_Termination/course_exit: 0.5214 Episode_Termination/illegal_nonwheel_contact: 0.0286 ============================================================================== ============================================================================== Learning iteration 41700/50000 Total steps: 170,803,200 Steps per second: 13,618 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1420.0 Actor loss: -9.5100 Critic loss: +0.0045 World model loss: +0.5750 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1753 Episode_Reward/track_angular_velocity: +53.5925 Episode_Reward/upright: +27.1967 Episode_Reward/leg_posture: -2.7386 Episode_Reward/standstill: -6.9225 Episode_Reward/action_rate_l2: -0.6007 Episode_Reward/joint_torques_l2: -0.1506 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0277 Episode_Reward/unnecessary_wheel_lift: -0.0930 Episode_Reward/rolling_leg_motion: -0.6150 Episode_Reward/obstacle_progress: +3.8831 Episode_Reward/wheel_body_intrusion: -0.2735 Episode_Reward/pair_body_intrusion: -0.2859 Curriculum/terrain_level_mean: +4.2092 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0947 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4810 Episode_Termination/fell_over: 0.0316 Episode_Termination/course_exit: 0.4557 Episode_Termination/illegal_nonwheel_contact: 0.0506 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2388] ============================================================================== Learning iteration 41750/50000 Total steps: 171,008,000 Steps per second: 13,620 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0870 Mean episode length: 1351.8 Actor loss: -9.4882 Critic loss: +0.0058 World model loss: +0.5995 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7983 Episode_Reward/track_angular_velocity: +51.0182 Episode_Reward/upright: +25.9016 Episode_Reward/leg_posture: -2.5930 Episode_Reward/standstill: -6.2286 Episode_Reward/action_rate_l2: -0.5721 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.0933 Episode_Reward/rolling_leg_motion: -0.5958 Episode_Reward/obstacle_progress: +3.7564 Episode_Reward/wheel_body_intrusion: -0.2688 Episode_Reward/pair_body_intrusion: -0.3036 Curriculum/terrain_level_mean: +4.2122 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0946 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4667 Episode_Termination/fell_over: 0.0364 Episode_Termination/course_exit: 0.4788 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== ============================================================================== Learning iteration 41800/50000 Total steps: 171,212,800 Steps per second: 13,622 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1407.4 Actor loss: -9.5263 Critic loss: +0.0045 World model loss: +0.5873 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8040 Episode_Reward/track_angular_velocity: +53.1420 Episode_Reward/upright: +26.9926 Episode_Reward/leg_posture: -2.6999 Episode_Reward/standstill: -6.5039 Episode_Reward/action_rate_l2: -0.5886 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0137 Episode_Reward/unnecessary_wheel_lift: -0.0892 Episode_Reward/rolling_leg_motion: -0.6176 Episode_Reward/obstacle_progress: +3.9285 Episode_Reward/wheel_body_intrusion: -0.2817 Episode_Reward/pair_body_intrusion: -0.3122 Curriculum/terrain_level_mean: +4.2144 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5316 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4367 Episode_Termination/illegal_nonwheel_contact: 0.0316 ============================================================================== ============================================================================== Learning iteration 41850/50000 Total steps: 171,417,600 Steps per second: 13,625 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1345.8 Actor loss: -9.5118 Critic loss: +0.0034 World model loss: +0.5360 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8578 Episode_Reward/track_angular_velocity: +50.8411 Episode_Reward/upright: +25.8267 Episode_Reward/leg_posture: -2.5790 Episode_Reward/standstill: -6.1167 Episode_Reward/action_rate_l2: -0.5597 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0131 Episode_Reward/unnecessary_wheel_lift: -0.0894 Episode_Reward/rolling_leg_motion: -0.5668 Episode_Reward/obstacle_progress: +3.8241 Episode_Reward/wheel_body_intrusion: -0.2686 Episode_Reward/pair_body_intrusion: -0.3051 Curriculum/terrain_level_mean: +4.2156 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4759 Episode_Termination/fell_over: 0.0207 Episode_Termination/course_exit: 0.4897 Episode_Termination/illegal_nonwheel_contact: 0.0138 ============================================================================== ============================================================================== Learning iteration 41900/50000 Total steps: 171,622,400 Steps per second: 13,627 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1473.1 Actor loss: -9.5014 Critic loss: +0.0047 World model loss: +0.5514 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +52.3044 Episode_Reward/track_angular_velocity: +55.6743 Episode_Reward/upright: +28.2521 Episode_Reward/leg_posture: -2.8187 Episode_Reward/standstill: -6.3227 Episode_Reward/action_rate_l2: -0.6093 Episode_Reward/joint_torques_l2: -0.1544 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0115 Episode_Reward/unnecessary_wheel_lift: -0.0866 Episode_Reward/rolling_leg_motion: -0.6345 Episode_Reward/obstacle_progress: +4.1235 Episode_Reward/wheel_body_intrusion: -0.2888 Episode_Reward/pair_body_intrusion: -0.3056 Curriculum/terrain_level_mean: +4.2170 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5563 Episode_Termination/fell_over: 0.0125 Episode_Termination/course_exit: 0.4188 Episode_Termination/illegal_nonwheel_contact: 0.0125 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[864] ============================================================================== Learning iteration 41950/50000 Total steps: 171,827,200 Steps per second: 13,629 Iteration time: 0.301s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1370.2 Actor loss: -9.4956 Critic loss: +0.0042 World model loss: +0.5817 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6836 Episode_Reward/track_angular_velocity: +51.7775 Episode_Reward/upright: +26.2567 Episode_Reward/leg_posture: -2.5886 Episode_Reward/standstill: -5.8891 Episode_Reward/action_rate_l2: -0.5779 Episode_Reward/joint_torques_l2: -0.1435 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0714 Episode_Reward/rolling_leg_motion: -0.5861 Episode_Reward/obstacle_progress: +3.8276 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.2848 Curriculum/terrain_level_mean: +4.2192 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4779 Episode_Termination/fell_over: 0.0074 Episode_Termination/course_exit: 0.4853 Episode_Termination/illegal_nonwheel_contact: 0.0294 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1114] ============================================================================== Learning iteration 42000/50000 Total steps: 172,032,000 Steps per second: 13,631 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1375.6 Actor loss: -9.5003 Critic loss: +0.0051 World model loss: +0.6024 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6976 Episode_Reward/track_angular_velocity: +51.9948 Episode_Reward/upright: +26.3727 Episode_Reward/leg_posture: -2.5922 Episode_Reward/standstill: -5.9703 Episode_Reward/action_rate_l2: -0.5886 Episode_Reward/joint_torques_l2: -0.1439 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0834 Episode_Reward/rolling_leg_motion: -0.5898 Episode_Reward/obstacle_progress: +3.7939 Episode_Reward/wheel_body_intrusion: -0.2627 Episode_Reward/pair_body_intrusion: -0.2955 Curriculum/terrain_level_mean: +4.2219 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4852 Episode_Termination/fell_over: 0.0178 Episode_Termination/course_exit: 0.4911 Episode_Termination/illegal_nonwheel_contact: 0.0178 ============================================================================== ============================================================================== Learning iteration 42050/50000 Total steps: 172,236,800 Steps per second: 13,633 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1315.4 Actor loss: -9.5001 Critic loss: +0.0057 World model loss: +0.5700 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5954 Episode_Reward/track_angular_velocity: +49.6777 Episode_Reward/upright: +25.1846 Episode_Reward/leg_posture: -2.5192 Episode_Reward/standstill: -5.7920 Episode_Reward/action_rate_l2: -0.5609 Episode_Reward/joint_torques_l2: -0.1383 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0231 Episode_Reward/unnecessary_wheel_lift: -0.0832 Episode_Reward/rolling_leg_motion: -0.5553 Episode_Reward/obstacle_progress: +3.6340 Episode_Reward/wheel_body_intrusion: -0.2519 Episode_Reward/pair_body_intrusion: -0.2824 Curriculum/terrain_level_mean: +4.2224 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0370 Episode_Termination/course_exit: 0.4889 Episode_Termination/illegal_nonwheel_contact: 0.0370 ============================================================================== ============================================================================== Learning iteration 42100/50000 Total steps: 172,441,600 Steps per second: 13,635 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1339.0 Actor loss: -9.5253 Critic loss: +0.0039 World model loss: +0.5587 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3851 Episode_Reward/track_angular_velocity: +50.5471 Episode_Reward/upright: +25.6344 Episode_Reward/leg_posture: -2.5249 Episode_Reward/standstill: -5.9683 Episode_Reward/action_rate_l2: -0.5820 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0897 Episode_Reward/rolling_leg_motion: -0.5609 Episode_Reward/obstacle_progress: +3.7742 Episode_Reward/wheel_body_intrusion: -0.2464 Episode_Reward/pair_body_intrusion: -0.2857 Curriculum/terrain_level_mean: +4.2246 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4417 Episode_Termination/fell_over: 0.0307 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0307 ============================================================================== ============================================================================== Learning iteration 42150/50000 Total steps: 172,646,400 Steps per second: 13,637 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1333.3 Actor loss: -9.5222 Critic loss: +0.0038 World model loss: +0.5798 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1637 Episode_Reward/track_angular_velocity: +50.2949 Episode_Reward/upright: +25.5068 Episode_Reward/leg_posture: -2.4985 Episode_Reward/standstill: -5.6352 Episode_Reward/action_rate_l2: -0.5743 Episode_Reward/joint_torques_l2: -0.1396 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0182 Episode_Reward/unnecessary_wheel_lift: -0.0700 Episode_Reward/rolling_leg_motion: -0.5506 Episode_Reward/obstacle_progress: +3.7707 Episode_Reward/wheel_body_intrusion: -0.2476 Episode_Reward/pair_body_intrusion: -0.2683 Curriculum/terrain_level_mean: +4.2258 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4853 Episode_Termination/fell_over: 0.0221 Episode_Termination/course_exit: 0.4632 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 42200/50000 Total steps: 172,851,200 Steps per second: 13,640 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1300.8 Actor loss: -9.4958 Critic loss: +0.0055 World model loss: +0.5655 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.9724 Episode_Reward/track_angular_velocity: +49.0495 Episode_Reward/upright: +24.8959 Episode_Reward/leg_posture: -2.4678 Episode_Reward/standstill: -5.7479 Episode_Reward/action_rate_l2: -0.5602 Episode_Reward/joint_torques_l2: -0.1372 Episode_Reward/dof_pos_limits: -0.0015 Episode_Reward/terminated: -0.0254 Episode_Reward/unnecessary_wheel_lift: -0.0861 Episode_Reward/rolling_leg_motion: -0.5538 Episode_Reward/obstacle_progress: +3.6659 Episode_Reward/wheel_body_intrusion: -0.2589 Episode_Reward/pair_body_intrusion: -0.2568 Curriculum/terrain_level_mean: +4.2271 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1370 Episode_Termination/time_out: 0.4277 Episode_Termination/fell_over: 0.0440 Episode_Termination/course_exit: 0.5094 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 42250/50000 Total steps: 173,056,000 Steps per second: 13,642 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1305.6 Actor loss: -9.5216 Critic loss: +0.0046 World model loss: +0.5553 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0783 Episode_Reward/track_angular_velocity: +49.3290 Episode_Reward/upright: +24.9736 Episode_Reward/leg_posture: -2.4542 Episode_Reward/standstill: -6.1558 Episode_Reward/action_rate_l2: -0.5672 Episode_Reward/joint_torques_l2: -0.1369 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0357 Episode_Reward/unnecessary_wheel_lift: -0.0804 Episode_Reward/rolling_leg_motion: -0.5392 Episode_Reward/obstacle_progress: +3.6858 Episode_Reward/wheel_body_intrusion: -0.2439 Episode_Reward/pair_body_intrusion: -0.2529 Curriculum/terrain_level_mean: +4.2258 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0953 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0523 Episode_Termination/course_exit: 0.4641 Episode_Termination/illegal_nonwheel_contact: 0.0588 ============================================================================== ============================================================================== Learning iteration 42300/50000 Total steps: 173,260,800 Steps per second: 13,644 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0870 Mean episode length: 1307.3 Actor loss: -9.5223 Critic loss: +0.0041 World model loss: +0.5697 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2427 Episode_Reward/track_angular_velocity: +49.3580 Episode_Reward/upright: +25.0064 Episode_Reward/leg_posture: -2.4622 Episode_Reward/standstill: -6.2734 Episode_Reward/action_rate_l2: -0.5669 Episode_Reward/joint_torques_l2: -0.1373 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0332 Episode_Reward/unnecessary_wheel_lift: -0.0949 Episode_Reward/rolling_leg_motion: -0.5511 Episode_Reward/obstacle_progress: +3.7123 Episode_Reward/wheel_body_intrusion: -0.2407 Episode_Reward/pair_body_intrusion: -0.2585 Curriculum/terrain_level_mean: +4.2241 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4387 Episode_Termination/fell_over: 0.0581 Episode_Termination/course_exit: 0.4839 Episode_Termination/illegal_nonwheel_contact: 0.0581 ============================================================================== ============================================================================== Learning iteration 42350/50000 Total steps: 173,465,600 Steps per second: 13,646 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0870 Mean episode length: 1370.3 Actor loss: -9.4991 Critic loss: +0.0039 World model loss: +0.5710 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5495 Episode_Reward/track_angular_velocity: +51.7270 Episode_Reward/upright: +26.2402 Episode_Reward/leg_posture: -2.5647 Episode_Reward/standstill: -6.6730 Episode_Reward/action_rate_l2: -0.5936 Episode_Reward/joint_torques_l2: -0.1441 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0916 Episode_Reward/rolling_leg_motion: -0.5794 Episode_Reward/obstacle_progress: +3.8961 Episode_Reward/wheel_body_intrusion: -0.2483 Episode_Reward/pair_body_intrusion: -0.2946 Curriculum/terrain_level_mean: +4.2253 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5094 Episode_Termination/fell_over: 0.0377 Episode_Termination/course_exit: 0.4465 Episode_Termination/illegal_nonwheel_contact: 0.0189 ============================================================================== ============================================================================== Learning iteration 42400/50000 Total steps: 173,670,400 Steps per second: 13,648 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1360.0 Actor loss: -9.5251 Critic loss: +0.0049 World model loss: +0.5874 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2230 Episode_Reward/track_angular_velocity: +51.4069 Episode_Reward/upright: +26.0752 Episode_Reward/leg_posture: -2.5544 Episode_Reward/standstill: -5.9104 Episode_Reward/action_rate_l2: -0.5851 Episode_Reward/joint_torques_l2: -0.1426 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.0956 Episode_Reward/rolling_leg_motion: -0.5808 Episode_Reward/obstacle_progress: +3.8457 Episode_Reward/wheel_body_intrusion: -0.2533 Episode_Reward/pair_body_intrusion: -0.2856 Curriculum/terrain_level_mean: +4.2263 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4634 Episode_Termination/fell_over: 0.0122 Episode_Termination/course_exit: 0.4878 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== ============================================================================== Learning iteration 42450/50000 Total steps: 173,875,200 Steps per second: 13,650 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1316.7 Actor loss: -9.5393 Critic loss: +0.0081 World model loss: +0.5883 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6891 Episode_Reward/track_angular_velocity: +49.7616 Episode_Reward/upright: +25.2314 Episode_Reward/leg_posture: -2.4927 Episode_Reward/standstill: -5.9936 Episode_Reward/action_rate_l2: -0.5603 Episode_Reward/joint_torques_l2: -0.1380 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0309 Episode_Reward/unnecessary_wheel_lift: -0.0882 Episode_Reward/rolling_leg_motion: -0.5522 Episode_Reward/obstacle_progress: +3.7122 Episode_Reward/wheel_body_intrusion: -0.2551 Episode_Reward/pair_body_intrusion: -0.2875 Curriculum/terrain_level_mean: +4.2261 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4533 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.4733 Episode_Termination/illegal_nonwheel_contact: 0.0733 ============================================================================== ============================================================================== Learning iteration 42500/50000 Total steps: 174,080,000 Steps per second: 13,652 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1359.5 Actor loss: -9.5180 Critic loss: +0.0071 World model loss: +0.5613 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.2640 Episode_Reward/track_angular_velocity: +51.3877 Episode_Reward/upright: +26.0497 Episode_Reward/leg_posture: -2.6147 Episode_Reward/standstill: -6.0752 Episode_Reward/action_rate_l2: -0.5751 Episode_Reward/joint_torques_l2: -0.1422 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.0845 Episode_Reward/rolling_leg_motion: -0.5769 Episode_Reward/obstacle_progress: +3.7899 Episode_Reward/wheel_body_intrusion: -0.2767 Episode_Reward/pair_body_intrusion: -0.2892 Curriculum/terrain_level_mean: +4.2280 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4516 Episode_Termination/fell_over: 0.0065 Episode_Termination/course_exit: 0.5097 Episode_Termination/illegal_nonwheel_contact: 0.0323 ============================================================================== ============================================================================== Learning iteration 42550/50000 Total steps: 174,284,800 Steps per second: 13,654 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1318.5 Actor loss: -9.4829 Critic loss: +0.0062 World model loss: +0.5810 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8873 Episode_Reward/track_angular_velocity: +49.7940 Episode_Reward/upright: +25.2393 Episode_Reward/leg_posture: -2.5515 Episode_Reward/standstill: -6.0929 Episode_Reward/action_rate_l2: -0.5674 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0754 Episode_Reward/rolling_leg_motion: -0.5722 Episode_Reward/obstacle_progress: +3.6629 Episode_Reward/wheel_body_intrusion: -0.2862 Episode_Reward/pair_body_intrusion: -0.2905 Curriculum/terrain_level_mean: +4.2285 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4091 Episode_Termination/fell_over: 0.0303 Episode_Termination/course_exit: 0.5303 Episode_Termination/illegal_nonwheel_contact: 0.0530 ============================================================================== ============================================================================== Learning iteration 42600/50000 Total steps: 174,489,600 Steps per second: 13,656 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0875 Mean episode length: 1372.0 Actor loss: -9.4970 Critic loss: +0.0086 World model loss: +0.5966 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6938 Episode_Reward/track_angular_velocity: +51.8237 Episode_Reward/upright: +26.3111 Episode_Reward/leg_posture: -2.5452 Episode_Reward/standstill: -6.0627 Episode_Reward/action_rate_l2: -0.5820 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0814 Episode_Reward/rolling_leg_motion: -0.5987 Episode_Reward/obstacle_progress: +3.8551 Episode_Reward/wheel_body_intrusion: -0.2844 Episode_Reward/pair_body_intrusion: -0.2662 Curriculum/terrain_level_mean: +4.2302 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.5210 Episode_Termination/fell_over: 0.0359 Episode_Termination/course_exit: 0.4371 Episode_Termination/illegal_nonwheel_contact: 0.0240 ============================================================================== ============================================================================== Learning iteration 42650/50000 Total steps: 174,694,400 Steps per second: 13,658 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1382.0 Actor loss: -9.4925 Critic loss: +0.0052 World model loss: +0.5723 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9931 Episode_Reward/track_angular_velocity: +52.1416 Episode_Reward/upright: +26.5210 Episode_Reward/leg_posture: -2.5997 Episode_Reward/standstill: -6.4374 Episode_Reward/action_rate_l2: -0.5941 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0297 Episode_Reward/unnecessary_wheel_lift: -0.0877 Episode_Reward/rolling_leg_motion: -0.5883 Episode_Reward/obstacle_progress: +3.9591 Episode_Reward/wheel_body_intrusion: -0.2767 Episode_Reward/pair_body_intrusion: -0.2711 Curriculum/terrain_level_mean: +4.2307 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4765 Episode_Termination/fell_over: 0.0537 Episode_Termination/course_exit: 0.4497 Episode_Termination/illegal_nonwheel_contact: 0.0403 ============================================================================== ============================================================================== Learning iteration 42700/50000 Total steps: 174,899,200 Steps per second: 13,660 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1353.5 Actor loss: -9.5136 Critic loss: +0.0052 World model loss: +0.5757 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0000 Episode_Reward/track_angular_velocity: +51.0752 Episode_Reward/upright: +25.9549 Episode_Reward/leg_posture: -2.5402 Episode_Reward/standstill: -5.9150 Episode_Reward/action_rate_l2: -0.5785 Episode_Reward/joint_torques_l2: -0.1432 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0747 Episode_Reward/rolling_leg_motion: -0.5644 Episode_Reward/obstacle_progress: +3.9016 Episode_Reward/wheel_body_intrusion: -0.2755 Episode_Reward/pair_body_intrusion: -0.2546 Curriculum/terrain_level_mean: +4.2310 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4785 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.4785 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 42750/50000 Total steps: 175,104,000 Steps per second: 13,663 Iteration time: 0.300s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0874 Mean episode length: 1366.3 Actor loss: -9.5420 Critic loss: +0.0039 World model loss: +0.5472 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4793 Episode_Reward/track_angular_velocity: +51.5689 Episode_Reward/upright: +26.1779 Episode_Reward/leg_posture: -2.5541 Episode_Reward/standstill: -6.1740 Episode_Reward/action_rate_l2: -0.5780 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0766 Episode_Reward/rolling_leg_motion: -0.5673 Episode_Reward/obstacle_progress: +3.8874 Episode_Reward/wheel_body_intrusion: -0.2875 Episode_Reward/pair_body_intrusion: -0.3144 Curriculum/terrain_level_mean: +4.2314 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4820 Episode_Termination/fell_over: 0.0072 Episode_Termination/course_exit: 0.4676 Episode_Termination/illegal_nonwheel_contact: 0.0504 ============================================================================== ============================================================================== Learning iteration 42800/50000 Total steps: 175,308,800 Steps per second: 13,665 Iteration time: 0.300s Collection time: 0.075s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1278.4 Actor loss: -9.5151 Critic loss: +0.0090 World model loss: +0.5912 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.3328 Episode_Reward/track_angular_velocity: +48.2986 Episode_Reward/upright: +24.4669 Episode_Reward/leg_posture: -2.3991 Episode_Reward/standstill: -6.0178 Episode_Reward/action_rate_l2: -0.5390 Episode_Reward/joint_torques_l2: -0.1354 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0200 Episode_Reward/unnecessary_wheel_lift: -0.0891 Episode_Reward/rolling_leg_motion: -0.5469 Episode_Reward/obstacle_progress: +3.6805 Episode_Reward/wheel_body_intrusion: -0.2546 Episode_Reward/pair_body_intrusion: -0.2697 Curriculum/terrain_level_mean: +4.2322 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4359 Episode_Termination/fell_over: 0.0192 Episode_Termination/course_exit: 0.5128 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1711] RECOVERY reset 1 invalid environments; first IDs=[588] ============================================================================== Learning iteration 42850/50000 Total steps: 175,513,600 Steps per second: 13,667 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1330.1 Actor loss: -9.4952 Critic loss: +0.0059 World model loss: +0.5908 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1802 Episode_Reward/track_angular_velocity: +50.1933 Episode_Reward/upright: +25.4626 Episode_Reward/leg_posture: -2.5286 Episode_Reward/standstill: -6.3852 Episode_Reward/action_rate_l2: -0.5690 Episode_Reward/joint_torques_l2: -0.1412 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0312 Episode_Reward/unnecessary_wheel_lift: -0.0962 Episode_Reward/rolling_leg_motion: -0.5759 Episode_Reward/obstacle_progress: +3.8250 Episode_Reward/wheel_body_intrusion: -0.2597 Episode_Reward/pair_body_intrusion: -0.2476 Curriculum/terrain_level_mean: +4.2317 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4479 Episode_Termination/fell_over: 0.0368 Episode_Termination/course_exit: 0.4785 Episode_Termination/illegal_nonwheel_contact: 0.0613 ============================================================================== ============================================================================== Learning iteration 42900/50000 Total steps: 175,718,400 Steps per second: 13,669 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0871 Mean episode length: 1399.5 Actor loss: -9.5362 Critic loss: +0.0072 World model loss: +0.5913 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.6698 Episode_Reward/track_angular_velocity: +52.8812 Episode_Reward/upright: +26.8105 Episode_Reward/leg_posture: -2.6404 Episode_Reward/standstill: -6.7738 Episode_Reward/action_rate_l2: -0.5895 Episode_Reward/joint_torques_l2: -0.1480 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0273 Episode_Reward/unnecessary_wheel_lift: -0.1021 Episode_Reward/rolling_leg_motion: -0.5977 Episode_Reward/obstacle_progress: +4.0167 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.2833 Curriculum/terrain_level_mean: +4.2322 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0473 Episode_Termination/course_exit: 0.4392 Episode_Termination/illegal_nonwheel_contact: 0.0541 ============================================================================== ============================================================================== Learning iteration 42950/50000 Total steps: 175,923,200 Steps per second: 13,671 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1370.7 Actor loss: -9.5343 Critic loss: +0.0037 World model loss: +0.5845 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6199 Episode_Reward/track_angular_velocity: +51.7085 Episode_Reward/upright: +26.2417 Episode_Reward/leg_posture: -2.5988 Episode_Reward/standstill: -6.1591 Episode_Reward/action_rate_l2: -0.5781 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0294 Episode_Reward/unnecessary_wheel_lift: -0.0861 Episode_Reward/rolling_leg_motion: -0.5738 Episode_Reward/obstacle_progress: +3.8985 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.2720 Curriculum/terrain_level_mean: +4.2319 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.5192 Episode_Termination/fell_over: 0.0256 Episode_Termination/course_exit: 0.4103 Episode_Termination/illegal_nonwheel_contact: 0.0641 ============================================================================== ============================================================================== Learning iteration 43000/50000 Total steps: 176,128,000 Steps per second: 13,673 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0869 Mean episode length: 1317.6 Actor loss: -9.5161 Critic loss: +0.0055 World model loss: +0.5443 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7318 Episode_Reward/track_angular_velocity: +49.6951 Episode_Reward/upright: +25.2117 Episode_Reward/leg_posture: -2.5221 Episode_Reward/standstill: -6.4274 Episode_Reward/action_rate_l2: -0.5480 Episode_Reward/joint_torques_l2: -0.1395 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0335 Episode_Reward/unnecessary_wheel_lift: -0.0952 Episode_Reward/rolling_leg_motion: -0.5491 Episode_Reward/obstacle_progress: +3.6844 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.2899 Curriculum/terrain_level_mean: +4.2317 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.3931 Episode_Termination/fell_over: 0.0690 Episode_Termination/course_exit: 0.5172 Episode_Termination/illegal_nonwheel_contact: 0.0552 ============================================================================== ============================================================================== Learning iteration 43050/50000 Total steps: 176,332,800 Steps per second: 13,675 Iteration time: 0.300s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1344.2 Actor loss: -9.5347 Critic loss: +0.0063 World model loss: +0.5965 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7929 Episode_Reward/track_angular_velocity: +50.7244 Episode_Reward/upright: +25.7647 Episode_Reward/leg_posture: -2.5357 Episode_Reward/standstill: -6.1141 Episode_Reward/action_rate_l2: -0.5592 Episode_Reward/joint_torques_l2: -0.1419 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0289 Episode_Reward/unnecessary_wheel_lift: -0.1001 Episode_Reward/rolling_leg_motion: -0.5645 Episode_Reward/obstacle_progress: +3.7562 Episode_Reward/wheel_body_intrusion: -0.2642 Episode_Reward/pair_body_intrusion: -0.3105 Curriculum/terrain_level_mean: +4.2322 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.4667 Episode_Termination/fell_over: 0.0485 Episode_Termination/course_exit: 0.4667 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[3356] RECOVERY reset 1 invalid environments; first IDs=[2338] ============================================================================== Learning iteration 43100/50000 Total steps: 176,537,600 Steps per second: 13,677 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1324.9 Actor loss: -9.5172 Critic loss: +0.0063 World model loss: +0.5616 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1451 Episode_Reward/track_angular_velocity: +50.0530 Episode_Reward/upright: +25.3959 Episode_Reward/leg_posture: -2.4809 Episode_Reward/standstill: -6.1673 Episode_Reward/action_rate_l2: -0.5506 Episode_Reward/joint_torques_l2: -0.1400 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.0894 Episode_Reward/rolling_leg_motion: -0.5594 Episode_Reward/obstacle_progress: +3.6825 Episode_Reward/wheel_body_intrusion: -0.2729 Episode_Reward/pair_body_intrusion: -0.2914 Curriculum/terrain_level_mean: +4.2344 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.4295 Episode_Termination/fell_over: 0.0064 Episode_Termination/course_exit: 0.5192 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== ============================================================================== Learning iteration 43150/50000 Total steps: 176,742,400 Steps per second: 13,679 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1396.5 Actor loss: -9.5265 Critic loss: +0.0049 World model loss: +0.5611 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7922 Episode_Reward/track_angular_velocity: +52.7108 Episode_Reward/upright: +26.7589 Episode_Reward/leg_posture: -2.6050 Episode_Reward/standstill: -6.3628 Episode_Reward/action_rate_l2: -0.5825 Episode_Reward/joint_torques_l2: -0.1477 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0136 Episode_Reward/unnecessary_wheel_lift: -0.0769 Episode_Reward/rolling_leg_motion: -0.5825 Episode_Reward/obstacle_progress: +3.9908 Episode_Reward/wheel_body_intrusion: -0.2889 Episode_Reward/pair_body_intrusion: -0.3203 Curriculum/terrain_level_mean: +4.2354 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.5390 Episode_Termination/fell_over: 0.0130 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0260 ============================================================================== ============================================================================== Learning iteration 43200/50000 Total steps: 176,947,200 Steps per second: 13,681 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1399.3 Actor loss: -9.5206 Critic loss: +0.0061 World model loss: +0.5617 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8252 Episode_Reward/track_angular_velocity: +52.8455 Episode_Reward/upright: +26.8108 Episode_Reward/leg_posture: -2.6430 Episode_Reward/standstill: -6.6750 Episode_Reward/action_rate_l2: -0.5782 Episode_Reward/joint_torques_l2: -0.1487 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.0868 Episode_Reward/rolling_leg_motion: -0.5751 Episode_Reward/obstacle_progress: +3.9770 Episode_Reward/wheel_body_intrusion: -0.3070 Episode_Reward/pair_body_intrusion: -0.3265 Curriculum/terrain_level_mean: +4.2356 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4961 Episode_Termination/fell_over: 0.0388 Episode_Termination/course_exit: 0.4496 Episode_Termination/illegal_nonwheel_contact: 0.0310 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_043200.mp4 ============================================================================== Learning iteration 43250/50000 Total steps: 177,152,000 Steps per second: 13,676 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1349.6 Actor loss: -9.5287 Critic loss: +0.0057 World model loss: +0.5392 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9900 Episode_Reward/track_angular_velocity: +50.9503 Episode_Reward/upright: +25.8439 Episode_Reward/leg_posture: -2.5939 Episode_Reward/standstill: -6.7425 Episode_Reward/action_rate_l2: -0.5578 Episode_Reward/joint_torques_l2: -0.1431 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0290 Episode_Reward/unnecessary_wheel_lift: -0.0826 Episode_Reward/rolling_leg_motion: -0.5621 Episode_Reward/obstacle_progress: +3.8675 Episode_Reward/wheel_body_intrusion: -0.2748 Episode_Reward/pair_body_intrusion: -0.3379 Curriculum/terrain_level_mean: +4.2356 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4776 Episode_Termination/fell_over: 0.0522 Episode_Termination/course_exit: 0.4403 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 43300/50000 Total steps: 177,356,800 Steps per second: 13,678 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1348.8 Actor loss: -9.5546 Critic loss: +0.0048 World model loss: +0.6049 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8998 Episode_Reward/track_angular_velocity: +50.9741 Episode_Reward/upright: +25.8575 Episode_Reward/leg_posture: -2.5801 Episode_Reward/standstill: -6.3354 Episode_Reward/action_rate_l2: -0.5530 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0847 Episode_Reward/rolling_leg_motion: -0.5667 Episode_Reward/obstacle_progress: +3.8404 Episode_Reward/wheel_body_intrusion: -0.2683 Episode_Reward/pair_body_intrusion: -0.2738 Curriculum/terrain_level_mean: +4.2363 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4211 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.5329 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== ============================================================================== Learning iteration 43350/50000 Total steps: 177,561,600 Steps per second: 13,680 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1335.5 Actor loss: -9.5337 Critic loss: +0.0046 World model loss: +0.5890 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5645 Episode_Reward/track_angular_velocity: +50.5122 Episode_Reward/upright: +25.5972 Episode_Reward/leg_posture: -2.5606 Episode_Reward/standstill: -6.0213 Episode_Reward/action_rate_l2: -0.5567 Episode_Reward/joint_torques_l2: -0.1436 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.0852 Episode_Reward/rolling_leg_motion: -0.5638 Episode_Reward/obstacle_progress: +3.8614 Episode_Reward/wheel_body_intrusion: -0.2687 Episode_Reward/pair_body_intrusion: -0.2848 Curriculum/terrain_level_mean: +4.2371 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4460 Episode_Termination/fell_over: 0.0216 Episode_Termination/course_exit: 0.5252 Episode_Termination/illegal_nonwheel_contact: 0.0216 ============================================================================== ============================================================================== Learning iteration 43400/50000 Total steps: 177,766,400 Steps per second: 13,682 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1360.3 Actor loss: -9.5303 Critic loss: +0.0044 World model loss: +0.5419 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.3402 Episode_Reward/track_angular_velocity: +51.3649 Episode_Reward/upright: +26.0772 Episode_Reward/leg_posture: -2.5871 Episode_Reward/standstill: -6.1949 Episode_Reward/action_rate_l2: -0.5633 Episode_Reward/joint_torques_l2: -0.1455 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0159 Episode_Reward/unnecessary_wheel_lift: -0.0789 Episode_Reward/rolling_leg_motion: -0.5768 Episode_Reward/obstacle_progress: +3.9792 Episode_Reward/wheel_body_intrusion: -0.2587 Episode_Reward/pair_body_intrusion: -0.2841 Curriculum/terrain_level_mean: +4.2380 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4672 Episode_Termination/fell_over: 0.0146 Episode_Termination/course_exit: 0.4891 Episode_Termination/illegal_nonwheel_contact: 0.0365 ============================================================================== ============================================================================== Learning iteration 43450/50000 Total steps: 177,971,200 Steps per second: 13,684 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1378.4 Actor loss: -9.5131 Critic loss: +0.0092 World model loss: +0.5889 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9042 Episode_Reward/track_angular_velocity: +52.0153 Episode_Reward/upright: +26.4383 Episode_Reward/leg_posture: -2.6346 Episode_Reward/standstill: -6.3666 Episode_Reward/action_rate_l2: -0.5631 Episode_Reward/joint_torques_l2: -0.1467 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0215 Episode_Reward/unnecessary_wheel_lift: -0.0894 Episode_Reward/rolling_leg_motion: -0.5764 Episode_Reward/obstacle_progress: +3.9292 Episode_Reward/wheel_body_intrusion: -0.2831 Episode_Reward/pair_body_intrusion: -0.2829 Curriculum/terrain_level_mean: +4.2393 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5033 Episode_Termination/fell_over: 0.0261 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2135] ============================================================================== Learning iteration 43500/50000 Total steps: 178,176,000 Steps per second: 13,686 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1323.4 Actor loss: -9.5469 Critic loss: +0.0040 World model loss: +0.5719 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1068 Episode_Reward/track_angular_velocity: +50.0289 Episode_Reward/upright: +25.3980 Episode_Reward/leg_posture: -2.4921 Episode_Reward/standstill: -6.2979 Episode_Reward/action_rate_l2: -0.5375 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0258 Episode_Reward/unnecessary_wheel_lift: -0.0930 Episode_Reward/rolling_leg_motion: -0.5634 Episode_Reward/obstacle_progress: +3.8477 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.2722 Curriculum/terrain_level_mean: +4.2397 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1338 Episode_Termination/time_out: 0.4540 Episode_Termination/fell_over: 0.0429 Episode_Termination/course_exit: 0.4847 Episode_Termination/illegal_nonwheel_contact: 0.0307 ============================================================================== ============================================================================== Learning iteration 43550/50000 Total steps: 178,380,800 Steps per second: 13,688 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0876 Mean episode length: 1390.8 Actor loss: -9.5155 Critic loss: +0.0074 World model loss: +0.5345 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3953 Episode_Reward/track_angular_velocity: +52.6116 Episode_Reward/upright: +26.7150 Episode_Reward/leg_posture: -2.6194 Episode_Reward/standstill: -6.3217 Episode_Reward/action_rate_l2: -0.5649 Episode_Reward/joint_torques_l2: -0.1464 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0054 Episode_Reward/unnecessary_wheel_lift: -0.0875 Episode_Reward/rolling_leg_motion: -0.5996 Episode_Reward/obstacle_progress: +3.9816 Episode_Reward/wheel_body_intrusion: -0.2594 Episode_Reward/pair_body_intrusion: -0.3008 Curriculum/terrain_level_mean: +4.2439 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4815 Episode_Termination/fell_over: 0.0062 Episode_Termination/course_exit: 0.5062 Episode_Termination/illegal_nonwheel_contact: 0.0062 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1931] ============================================================================== Learning iteration 43600/50000 Total steps: 178,585,600 Steps per second: 13,690 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1305.4 Actor loss: -9.5249 Critic loss: +0.0053 World model loss: +0.5865 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4999 Episode_Reward/track_angular_velocity: +49.4496 Episode_Reward/upright: +25.0540 Episode_Reward/leg_posture: -2.4339 Episode_Reward/standstill: -6.0047 Episode_Reward/action_rate_l2: -0.5438 Episode_Reward/joint_torques_l2: -0.1385 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0076 Episode_Reward/unnecessary_wheel_lift: -0.0699 Episode_Reward/rolling_leg_motion: -0.5561 Episode_Reward/obstacle_progress: +3.7793 Episode_Reward/wheel_body_intrusion: -0.2370 Episode_Reward/pair_body_intrusion: -0.2603 Curriculum/terrain_level_mean: +4.2458 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1316 Episode_Termination/time_out: 0.4143 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5643 Episode_Termination/illegal_nonwheel_contact: 0.0214 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1714] ============================================================================== Learning iteration 43650/50000 Total steps: 178,790,400 Steps per second: 13,692 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1423.8 Actor loss: -9.5326 Critic loss: +0.0034 World model loss: +0.5638 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.7512 Episode_Reward/track_angular_velocity: +53.8371 Episode_Reward/upright: +27.3374 Episode_Reward/leg_posture: -2.6531 Episode_Reward/standstill: -6.4411 Episode_Reward/action_rate_l2: -0.5976 Episode_Reward/joint_torques_l2: -0.1527 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0139 Episode_Reward/unnecessary_wheel_lift: -0.0786 Episode_Reward/rolling_leg_motion: -0.6059 Episode_Reward/obstacle_progress: +4.0902 Episode_Reward/wheel_body_intrusion: -0.2875 Episode_Reward/pair_body_intrusion: -0.3073 Curriculum/terrain_level_mean: +4.2466 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5782 Episode_Termination/fell_over: 0.0204 Episode_Termination/course_exit: 0.3878 Episode_Termination/illegal_nonwheel_contact: 0.0136 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2004] ============================================================================== Learning iteration 43700/50000 Total steps: 178,995,200 Steps per second: 13,694 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0876 Mean episode length: 1365.0 Actor loss: -9.5433 Critic loss: +0.0087 World model loss: +0.5909 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5472 Episode_Reward/track_angular_velocity: +51.5861 Episode_Reward/upright: +26.1948 Episode_Reward/leg_posture: -2.5476 Episode_Reward/standstill: -6.1248 Episode_Reward/action_rate_l2: -0.5748 Episode_Reward/joint_torques_l2: -0.1454 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0107 Episode_Reward/unnecessary_wheel_lift: -0.0770 Episode_Reward/rolling_leg_motion: -0.5805 Episode_Reward/obstacle_progress: +3.8801 Episode_Reward/wheel_body_intrusion: -0.2805 Episode_Reward/pair_body_intrusion: -0.3010 Curriculum/terrain_level_mean: +4.2488 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4962 Episode_Termination/fell_over: 0.0226 Episode_Termination/course_exit: 0.4737 Episode_Termination/illegal_nonwheel_contact: 0.0150 ============================================================================== ============================================================================== Learning iteration 43750/50000 Total steps: 179,200,000 Steps per second: 13,696 Iteration time: 0.299s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1411.4 Actor loss: -9.5203 Critic loss: +0.0073 World model loss: +0.5948 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.2442 Episode_Reward/track_angular_velocity: +53.3423 Episode_Reward/upright: +27.0440 Episode_Reward/leg_posture: -2.6407 Episode_Reward/standstill: -6.1385 Episode_Reward/action_rate_l2: -0.6058 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.1001 Episode_Reward/rolling_leg_motion: -0.6142 Episode_Reward/obstacle_progress: +4.0650 Episode_Reward/wheel_body_intrusion: -0.2585 Episode_Reward/pair_body_intrusion: -0.3098 Curriculum/terrain_level_mean: +4.2522 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.4717 Episode_Termination/illegal_nonwheel_contact: 0.0189 ============================================================================== ============================================================================== Learning iteration 43800/50000 Total steps: 179,404,800 Steps per second: 13,698 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1278.5 Actor loss: -9.5302 Critic loss: +0.0060 World model loss: +0.5508 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.5512 Episode_Reward/track_angular_velocity: +48.3491 Episode_Reward/upright: +24.5245 Episode_Reward/leg_posture: -2.3997 Episode_Reward/standstill: -5.7232 Episode_Reward/action_rate_l2: -0.5408 Episode_Reward/joint_torques_l2: -0.1355 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0325 Episode_Reward/unnecessary_wheel_lift: -0.0703 Episode_Reward/rolling_leg_motion: -0.5470 Episode_Reward/obstacle_progress: +3.6007 Episode_Reward/wheel_body_intrusion: -0.2457 Episode_Reward/pair_body_intrusion: -0.2572 Curriculum/terrain_level_mean: +4.2512 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0911 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.3830 Episode_Termination/fell_over: 0.0567 Episode_Termination/course_exit: 0.5319 Episode_Termination/illegal_nonwheel_contact: 0.0567 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2445] ============================================================================== Learning iteration 43850/50000 Total steps: 179,609,600 Steps per second: 13,700 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1373.0 Actor loss: -9.5481 Critic loss: +0.0035 World model loss: +0.5644 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8485 Episode_Reward/track_angular_velocity: +51.9013 Episode_Reward/upright: +26.3273 Episode_Reward/leg_posture: -2.6408 Episode_Reward/standstill: -6.1967 Episode_Reward/action_rate_l2: -0.5790 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0764 Episode_Reward/rolling_leg_motion: -0.5860 Episode_Reward/obstacle_progress: +3.7924 Episode_Reward/wheel_body_intrusion: -0.2786 Episode_Reward/pair_body_intrusion: -0.3153 Curriculum/terrain_level_mean: +4.2522 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4867 Episode_Termination/fell_over: 0.0267 Episode_Termination/course_exit: 0.4800 Episode_Termination/illegal_nonwheel_contact: 0.0200 ============================================================================== ============================================================================== Learning iteration 43900/50000 Total steps: 179,814,400 Steps per second: 13,702 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1295.9 Actor loss: -9.5260 Critic loss: +0.0039 World model loss: +0.5442 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1144 Episode_Reward/track_angular_velocity: +48.9485 Episode_Reward/upright: +24.8570 Episode_Reward/leg_posture: -2.5190 Episode_Reward/standstill: -5.7157 Episode_Reward/action_rate_l2: -0.5454 Episode_Reward/joint_torques_l2: -0.1381 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0152 Episode_Reward/unnecessary_wheel_lift: -0.0831 Episode_Reward/rolling_leg_motion: -0.5600 Episode_Reward/obstacle_progress: +3.6462 Episode_Reward/wheel_body_intrusion: -0.2730 Episode_Reward/pair_body_intrusion: -0.2731 Curriculum/terrain_level_mean: +4.2542 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4110 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.5479 Episode_Termination/illegal_nonwheel_contact: 0.0205 ============================================================================== ============================================================================== Learning iteration 43950/50000 Total steps: 180,019,200 Steps per second: 13,704 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1350.6 Actor loss: -9.5123 Critic loss: +0.0090 World model loss: +0.5489 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0740 Episode_Reward/track_angular_velocity: +51.0257 Episode_Reward/upright: +25.8972 Episode_Reward/leg_posture: -2.6047 Episode_Reward/standstill: -6.1521 Episode_Reward/action_rate_l2: -0.5784 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0317 Episode_Reward/unnecessary_wheel_lift: -0.0869 Episode_Reward/rolling_leg_motion: -0.5852 Episode_Reward/obstacle_progress: +3.8258 Episode_Reward/wheel_body_intrusion: -0.2632 Episode_Reward/pair_body_intrusion: -0.2832 Curriculum/terrain_level_mean: +4.2542 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4832 Episode_Termination/fell_over: 0.0537 Episode_Termination/course_exit: 0.4362 Episode_Termination/illegal_nonwheel_contact: 0.0604 ============================================================================== ============================================================================== Learning iteration 44000/50000 Total steps: 180,224,000 Steps per second: 13,706 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1391.5 Actor loss: -9.5158 Critic loss: +0.0063 World model loss: +0.5775 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5143 Episode_Reward/track_angular_velocity: +52.6240 Episode_Reward/upright: +26.6780 Episode_Reward/leg_posture: -2.7015 Episode_Reward/standstill: -6.4016 Episode_Reward/action_rate_l2: -0.5872 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0136 Episode_Reward/unnecessary_wheel_lift: -0.0712 Episode_Reward/rolling_leg_motion: -0.5813 Episode_Reward/obstacle_progress: +3.8756 Episode_Reward/wheel_body_intrusion: -0.3020 Episode_Reward/pair_body_intrusion: -0.2976 Curriculum/terrain_level_mean: +4.2561 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.4648 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== ============================================================================== Learning iteration 44050/50000 Total steps: 180,428,800 Steps per second: 13,708 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1397.5 Actor loss: -9.5177 Critic loss: +0.0057 World model loss: +0.5376 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5919 Episode_Reward/track_angular_velocity: +52.8259 Episode_Reward/upright: +26.7928 Episode_Reward/leg_posture: -2.6691 Episode_Reward/standstill: -6.1548 Episode_Reward/action_rate_l2: -0.5992 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0218 Episode_Reward/unnecessary_wheel_lift: -0.0817 Episode_Reward/rolling_leg_motion: -0.6051 Episode_Reward/obstacle_progress: +3.8482 Episode_Reward/wheel_body_intrusion: -0.2911 Episode_Reward/pair_body_intrusion: -0.2684 Curriculum/terrain_level_mean: +4.2563 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.5170 Episode_Termination/fell_over: 0.0068 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[968] ============================================================================== Learning iteration 44100/50000 Total steps: 180,633,600 Steps per second: 13,710 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1383.6 Actor loss: -9.5217 Critic loss: +0.0051 World model loss: +0.5477 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0514 Episode_Reward/track_angular_velocity: +52.2811 Episode_Reward/upright: +26.5133 Episode_Reward/leg_posture: -2.6877 Episode_Reward/standstill: -6.3695 Episode_Reward/action_rate_l2: -0.5939 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0867 Episode_Reward/rolling_leg_motion: -0.6132 Episode_Reward/obstacle_progress: +3.7620 Episode_Reward/wheel_body_intrusion: -0.2834 Episode_Reward/pair_body_intrusion: -0.2651 Curriculum/terrain_level_mean: +4.2581 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.4785 Episode_Termination/illegal_nonwheel_contact: 0.0245 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[914] RECOVERY reset 1 invalid environments; first IDs=[2413] ============================================================================== Learning iteration 44150/50000 Total steps: 180,838,400 Steps per second: 13,712 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1370.1 Actor loss: -9.5036 Critic loss: +0.0103 World model loss: +0.5815 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8004 Episode_Reward/track_angular_velocity: +51.8122 Episode_Reward/upright: +26.2713 Episode_Reward/leg_posture: -2.7017 Episode_Reward/standstill: -6.4398 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.1057 Episode_Reward/rolling_leg_motion: -0.6028 Episode_Reward/obstacle_progress: +3.7298 Episode_Reward/wheel_body_intrusion: -0.2656 Episode_Reward/pair_body_intrusion: -0.2974 Curriculum/terrain_level_mean: +4.2585 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4689 Episode_Termination/fell_over: 0.0282 Episode_Termination/course_exit: 0.4859 Episode_Termination/illegal_nonwheel_contact: 0.0282 ============================================================================== ============================================================================== Learning iteration 44200/50000 Total steps: 181,043,200 Steps per second: 13,714 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1371.3 Actor loss: -9.5385 Critic loss: +0.0043 World model loss: +0.5742 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8747 Episode_Reward/track_angular_velocity: +51.8224 Episode_Reward/upright: +26.3064 Episode_Reward/leg_posture: -2.6626 Episode_Reward/standstill: -6.4350 Episode_Reward/action_rate_l2: -0.5804 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0321 Episode_Reward/unnecessary_wheel_lift: -0.0877 Episode_Reward/rolling_leg_motion: -0.6111 Episode_Reward/obstacle_progress: +3.7729 Episode_Reward/wheel_body_intrusion: -0.2813 Episode_Reward/pair_body_intrusion: -0.2484 Curriculum/terrain_level_mean: +4.2581 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0909 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4969 Episode_Termination/fell_over: 0.0497 Episode_Termination/course_exit: 0.4286 Episode_Termination/illegal_nonwheel_contact: 0.0497 ============================================================================== ============================================================================== Learning iteration 44250/50000 Total steps: 181,248,000 Steps per second: 13,716 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0871 Mean episode length: 1410.5 Actor loss: -9.5524 Critic loss: +0.0048 World model loss: +0.5326 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1968 Episode_Reward/track_angular_velocity: +53.3274 Episode_Reward/upright: +27.0345 Episode_Reward/leg_posture: -2.7322 Episode_Reward/standstill: -6.7753 Episode_Reward/action_rate_l2: -0.5892 Episode_Reward/joint_torques_l2: -0.1517 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.1015 Episode_Reward/rolling_leg_motion: -0.5958 Episode_Reward/obstacle_progress: +3.8993 Episode_Reward/wheel_body_intrusion: -0.3005 Episode_Reward/pair_body_intrusion: -0.2975 Curriculum/terrain_level_mean: +4.2593 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0917 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5442 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4218 Episode_Termination/illegal_nonwheel_contact: 0.0136 ============================================================================== ============================================================================== Learning iteration 44300/50000 Total steps: 181,452,800 Steps per second: 13,718 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1322.9 Actor loss: -9.5085 Critic loss: +0.0068 World model loss: +0.5632 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0394 Episode_Reward/track_angular_velocity: +49.9706 Episode_Reward/upright: +25.3035 Episode_Reward/leg_posture: -2.5648 Episode_Reward/standstill: -5.8567 Episode_Reward/action_rate_l2: -0.5554 Episode_Reward/joint_torques_l2: -0.1393 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0310 Episode_Reward/unnecessary_wheel_lift: -0.0738 Episode_Reward/rolling_leg_motion: -0.5657 Episode_Reward/obstacle_progress: +3.6466 Episode_Reward/wheel_body_intrusion: -0.2750 Episode_Reward/pair_body_intrusion: -0.2750 Curriculum/terrain_level_mean: +4.2590 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.3873 Episode_Termination/fell_over: 0.0493 Episode_Termination/course_exit: 0.5282 Episode_Termination/illegal_nonwheel_contact: 0.0563 ============================================================================== ============================================================================== Learning iteration 44350/50000 Total steps: 181,657,600 Steps per second: 13,720 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.190s Mean reward: +0.0874 Mean episode length: 1370.8 Actor loss: -9.5364 Critic loss: +0.0041 World model loss: +0.5817 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7607 Episode_Reward/track_angular_velocity: +51.7710 Episode_Reward/upright: +26.2000 Episode_Reward/leg_posture: -2.6150 Episode_Reward/standstill: -6.2789 Episode_Reward/action_rate_l2: -0.5819 Episode_Reward/joint_torques_l2: -0.1448 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0187 Episode_Reward/unnecessary_wheel_lift: -0.0734 Episode_Reward/rolling_leg_motion: -0.5719 Episode_Reward/obstacle_progress: +3.8858 Episode_Reward/wheel_body_intrusion: -0.2704 Episode_Reward/pair_body_intrusion: -0.2697 Curriculum/terrain_level_mean: +4.2588 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.5109 Episode_Termination/fell_over: 0.0292 Episode_Termination/course_exit: 0.4380 Episode_Termination/illegal_nonwheel_contact: 0.0438 ============================================================================== ============================================================================== Learning iteration 44400/50000 Total steps: 181,862,400 Steps per second: 13,722 Iteration time: 0.299s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1354.0 Actor loss: -9.5115 Critic loss: +0.0053 World model loss: +0.5556 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0577 Episode_Reward/track_angular_velocity: +51.0971 Episode_Reward/upright: +25.9043 Episode_Reward/leg_posture: -2.5381 Episode_Reward/standstill: -6.4904 Episode_Reward/action_rate_l2: -0.5919 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0840 Episode_Reward/rolling_leg_motion: -0.5690 Episode_Reward/obstacle_progress: +3.8309 Episode_Reward/wheel_body_intrusion: -0.2574 Episode_Reward/pair_body_intrusion: -0.3013 Curriculum/terrain_level_mean: +4.2595 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4172 Episode_Termination/fell_over: 0.0123 Episode_Termination/course_exit: 0.5337 Episode_Termination/illegal_nonwheel_contact: 0.0368 ============================================================================== ============================================================================== Learning iteration 44450/50000 Total steps: 182,067,200 Steps per second: 13,724 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1366.3 Actor loss: -9.5096 Critic loss: +0.0038 World model loss: +0.5646 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4860 Episode_Reward/track_angular_velocity: +51.5488 Episode_Reward/upright: +26.1504 Episode_Reward/leg_posture: -2.5633 Episode_Reward/standstill: -6.3615 Episode_Reward/action_rate_l2: -0.5878 Episode_Reward/joint_torques_l2: -0.1455 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.0807 Episode_Reward/rolling_leg_motion: -0.5697 Episode_Reward/obstacle_progress: +3.8710 Episode_Reward/wheel_body_intrusion: -0.2723 Episode_Reward/pair_body_intrusion: -0.3137 Curriculum/terrain_level_mean: +4.2600 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4545 Episode_Termination/fell_over: 0.0364 Episode_Termination/course_exit: 0.4970 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== ============================================================================== Learning iteration 44500/50000 Total steps: 182,272,000 Steps per second: 13,726 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1344.0 Actor loss: -9.5417 Critic loss: +0.0057 World model loss: +0.5150 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7037 Episode_Reward/track_angular_velocity: +50.7171 Episode_Reward/upright: +25.7267 Episode_Reward/leg_posture: -2.5448 Episode_Reward/standstill: -6.2658 Episode_Reward/action_rate_l2: -0.5725 Episode_Reward/joint_torques_l2: -0.1424 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0131 Episode_Reward/unnecessary_wheel_lift: -0.0692 Episode_Reward/rolling_leg_motion: -0.5506 Episode_Reward/obstacle_progress: +3.8256 Episode_Reward/wheel_body_intrusion: -0.2729 Episode_Reward/pair_body_intrusion: -0.2622 Curriculum/terrain_level_mean: +4.2610 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4745 Episode_Termination/fell_over: 0.0219 Episode_Termination/course_exit: 0.4891 Episode_Termination/illegal_nonwheel_contact: 0.0146 ============================================================================== ============================================================================== Learning iteration 44550/50000 Total steps: 182,476,800 Steps per second: 13,728 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1318.7 Actor loss: -9.4954 Critic loss: +0.0094 World model loss: +0.5664 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6696 Episode_Reward/track_angular_velocity: +49.7636 Episode_Reward/upright: +25.2453 Episode_Reward/leg_posture: -2.5041 Episode_Reward/standstill: -6.6118 Episode_Reward/action_rate_l2: -0.5579 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0834 Episode_Reward/rolling_leg_motion: -0.5506 Episode_Reward/obstacle_progress: +3.7156 Episode_Reward/wheel_body_intrusion: -0.2719 Episode_Reward/pair_body_intrusion: -0.2739 Curriculum/terrain_level_mean: +4.2625 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4184 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.5390 Episode_Termination/illegal_nonwheel_contact: 0.0355 ============================================================================== ============================================================================== Learning iteration 44600/50000 Total steps: 182,681,600 Steps per second: 13,730 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0865 Mean episode length: 1372.5 Actor loss: -9.5411 Critic loss: +0.0041 World model loss: +0.5384 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6951 Episode_Reward/track_angular_velocity: +51.7677 Episode_Reward/upright: +26.3056 Episode_Reward/leg_posture: -2.6945 Episode_Reward/standstill: -7.0255 Episode_Reward/action_rate_l2: -0.5623 Episode_Reward/joint_torques_l2: -0.1478 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0295 Episode_Reward/unnecessary_wheel_lift: -0.1000 Episode_Reward/rolling_leg_motion: -0.5805 Episode_Reward/obstacle_progress: +3.7511 Episode_Reward/wheel_body_intrusion: -0.2972 Episode_Reward/pair_body_intrusion: -0.3018 Curriculum/terrain_level_mean: +4.2612 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4824 Episode_Termination/fell_over: 0.0412 Episode_Termination/course_exit: 0.4529 Episode_Termination/illegal_nonwheel_contact: 0.0412 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[743] ============================================================================== Learning iteration 44650/50000 Total steps: 182,886,400 Steps per second: 13,732 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0864 Mean episode length: 1347.2 Actor loss: -9.5362 Critic loss: +0.0042 World model loss: +0.5617 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7933 Episode_Reward/track_angular_velocity: +50.9354 Episode_Reward/upright: +25.8192 Episode_Reward/leg_posture: -2.6171 Episode_Reward/standstill: -7.1568 Episode_Reward/action_rate_l2: -0.5602 Episode_Reward/joint_torques_l2: -0.1425 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0395 Episode_Reward/unnecessary_wheel_lift: -0.1036 Episode_Reward/rolling_leg_motion: -0.5779 Episode_Reward/obstacle_progress: +3.6629 Episode_Reward/wheel_body_intrusion: -0.2819 Episode_Reward/pair_body_intrusion: -0.2963 Curriculum/terrain_level_mean: +4.2615 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4659 Episode_Termination/fell_over: 0.0511 Episode_Termination/course_exit: 0.4489 Episode_Termination/illegal_nonwheel_contact: 0.0625 ============================================================================== ============================================================================== Learning iteration 44700/50000 Total steps: 183,091,200 Steps per second: 13,734 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0868 Mean episode length: 1408.9 Actor loss: -9.5076 Critic loss: +0.0099 World model loss: +0.5677 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0269 Episode_Reward/track_angular_velocity: +53.2515 Episode_Reward/upright: +27.0151 Episode_Reward/leg_posture: -2.7675 Episode_Reward/standstill: -7.1327 Episode_Reward/action_rate_l2: -0.5800 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0220 Episode_Reward/unnecessary_wheel_lift: -0.0785 Episode_Reward/rolling_leg_motion: -0.5953 Episode_Reward/obstacle_progress: +3.8656 Episode_Reward/wheel_body_intrusion: -0.2947 Episode_Reward/pair_body_intrusion: -0.2803 Curriculum/terrain_level_mean: +4.2610 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0278 Episode_Termination/course_exit: 0.4444 Episode_Termination/illegal_nonwheel_contact: 0.0556 ============================================================================== ============================================================================== Learning iteration 44750/50000 Total steps: 183,296,000 Steps per second: 13,736 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1408.7 Actor loss: -9.5107 Critic loss: +0.0061 World model loss: +0.5668 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8968 Episode_Reward/track_angular_velocity: +53.2226 Episode_Reward/upright: +26.9965 Episode_Reward/leg_posture: -2.7969 Episode_Reward/standstill: -6.6435 Episode_Reward/action_rate_l2: -0.5789 Episode_Reward/joint_torques_l2: -0.1519 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0220 Episode_Reward/unnecessary_wheel_lift: -0.0848 Episode_Reward/rolling_leg_motion: -0.5791 Episode_Reward/obstacle_progress: +3.8612 Episode_Reward/wheel_body_intrusion: -0.3130 Episode_Reward/pair_body_intrusion: -0.3322 Curriculum/terrain_level_mean: +4.2600 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5573 Episode_Termination/fell_over: 0.0458 Episode_Termination/course_exit: 0.3817 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 44800/50000 Total steps: 183,500,800 Steps per second: 13,737 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1412.2 Actor loss: -9.5101 Critic loss: +0.0079 World model loss: +0.5417 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1040 Episode_Reward/track_angular_velocity: +53.3446 Episode_Reward/upright: +27.0901 Episode_Reward/leg_posture: -2.7717 Episode_Reward/standstill: -6.6038 Episode_Reward/action_rate_l2: -0.5791 Episode_Reward/joint_torques_l2: -0.1511 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0276 Episode_Reward/unnecessary_wheel_lift: -0.0795 Episode_Reward/rolling_leg_motion: -0.5881 Episode_Reward/obstacle_progress: +3.9442 Episode_Reward/wheel_body_intrusion: -0.3046 Episode_Reward/pair_body_intrusion: -0.3167 Curriculum/terrain_level_mean: +4.2603 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5197 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.4145 Episode_Termination/illegal_nonwheel_contact: 0.0658 ============================================================================== ============================================================================== Learning iteration 44850/50000 Total steps: 183,705,600 Steps per second: 13,739 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0874 Mean episode length: 1343.9 Actor loss: -9.4946 Critic loss: +0.0061 World model loss: +0.5391 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7801 Episode_Reward/track_angular_velocity: +50.8234 Episode_Reward/upright: +25.7715 Episode_Reward/leg_posture: -2.6114 Episode_Reward/standstill: -6.1258 Episode_Reward/action_rate_l2: -0.5540 Episode_Reward/joint_torques_l2: -0.1447 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0765 Episode_Reward/rolling_leg_motion: -0.5702 Episode_Reward/obstacle_progress: +3.7520 Episode_Reward/wheel_body_intrusion: -0.2866 Episode_Reward/pair_body_intrusion: -0.2961 Curriculum/terrain_level_mean: +4.2610 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1330 Episode_Termination/time_out: 0.4500 Episode_Termination/fell_over: 0.0143 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0429 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2230] ============================================================================== Learning iteration 44900/50000 Total steps: 183,910,400 Steps per second: 13,741 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0873 Mean episode length: 1398.8 Actor loss: -9.5016 Critic loss: +0.0045 World model loss: +0.5616 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5622 Episode_Reward/track_angular_velocity: +52.8849 Episode_Reward/upright: +26.8225 Episode_Reward/leg_posture: -2.7178 Episode_Reward/standstill: -6.2921 Episode_Reward/action_rate_l2: -0.5777 Episode_Reward/joint_torques_l2: -0.1509 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0908 Episode_Reward/rolling_leg_motion: -0.5852 Episode_Reward/obstacle_progress: +3.8489 Episode_Reward/wheel_body_intrusion: -0.2863 Episode_Reward/pair_body_intrusion: -0.2809 Curriculum/terrain_level_mean: +4.2620 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.5168 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.4564 Episode_Termination/illegal_nonwheel_contact: 0.0134 ============================================================================== ============================================================================== Learning iteration 44950/50000 Total steps: 184,115,200 Steps per second: 13,743 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0870 Mean episode length: 1326.1 Actor loss: -9.4694 Critic loss: +0.0040 World model loss: +0.6058 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1111 Episode_Reward/track_angular_velocity: +50.1101 Episode_Reward/upright: +25.4218 Episode_Reward/leg_posture: -2.5798 Episode_Reward/standstill: -6.4348 Episode_Reward/action_rate_l2: -0.5469 Episode_Reward/joint_torques_l2: -0.1418 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0337 Episode_Reward/unnecessary_wheel_lift: -0.0772 Episode_Reward/rolling_leg_motion: -0.5714 Episode_Reward/obstacle_progress: +3.6766 Episode_Reward/wheel_body_intrusion: -0.2753 Episode_Reward/pair_body_intrusion: -0.2554 Curriculum/terrain_level_mean: +4.2615 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1341 Episode_Termination/time_out: 0.4552 Episode_Termination/fell_over: 0.0621 Episode_Termination/course_exit: 0.4552 Episode_Termination/illegal_nonwheel_contact: 0.0621 ============================================================================== ============================================================================== Learning iteration 45000/50000 Total steps: 184,320,000 Steps per second: 13,745 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0870 Mean episode length: 1299.1 Actor loss: -9.4960 Critic loss: +0.0064 World model loss: +0.5567 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.0712 Episode_Reward/track_angular_velocity: +49.1251 Episode_Reward/upright: +24.9214 Episode_Reward/leg_posture: -2.4701 Episode_Reward/standstill: -6.2932 Episode_Reward/action_rate_l2: -0.5445 Episode_Reward/joint_torques_l2: -0.1391 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0304 Episode_Reward/unnecessary_wheel_lift: -0.0813 Episode_Reward/rolling_leg_motion: -0.5595 Episode_Reward/obstacle_progress: +3.5421 Episode_Reward/wheel_body_intrusion: -0.2928 Episode_Reward/pair_body_intrusion: -0.2826 Curriculum/terrain_level_mean: +4.2620 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4132 Episode_Termination/fell_over: 0.0299 Episode_Termination/course_exit: 0.5150 Episode_Termination/illegal_nonwheel_contact: 0.0479 ============================================================================== ============================================================================== Learning iteration 45050/50000 Total steps: 184,524,800 Steps per second: 13,746 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1304.3 Actor loss: -9.5193 Critic loss: +0.0049 World model loss: +0.5473 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2716 Episode_Reward/track_angular_velocity: +49.3218 Episode_Reward/upright: +25.0367 Episode_Reward/leg_posture: -2.4581 Episode_Reward/standstill: -6.2482 Episode_Reward/action_rate_l2: -0.5407 Episode_Reward/joint_torques_l2: -0.1385 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0127 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5410 Episode_Reward/obstacle_progress: +3.5964 Episode_Reward/wheel_body_intrusion: -0.2759 Episode_Reward/pair_body_intrusion: -0.2843 Curriculum/terrain_level_mean: +4.2639 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1334 Episode_Termination/time_out: 0.4384 Episode_Termination/fell_over: 0.0205 Episode_Termination/course_exit: 0.5274 Episode_Termination/illegal_nonwheel_contact: 0.0205 ============================================================================== ============================================================================== Learning iteration 45100/50000 Total steps: 184,729,600 Steps per second: 13,748 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1320.6 Actor loss: -9.5120 Critic loss: +0.0038 World model loss: +0.5485 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7929 Episode_Reward/track_angular_velocity: +49.8755 Episode_Reward/upright: +25.3589 Episode_Reward/leg_posture: -2.5192 Episode_Reward/standstill: -6.2204 Episode_Reward/action_rate_l2: -0.5535 Episode_Reward/joint_torques_l2: -0.1406 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0803 Episode_Reward/rolling_leg_motion: -0.5439 Episode_Reward/obstacle_progress: +3.7380 Episode_Reward/wheel_body_intrusion: -0.2801 Episode_Reward/pair_body_intrusion: -0.3014 Curriculum/terrain_level_mean: +4.2632 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4460 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.4964 Episode_Termination/illegal_nonwheel_contact: 0.0360 ============================================================================== ============================================================================== Learning iteration 45150/50000 Total steps: 184,934,400 Steps per second: 13,750 Iteration time: 0.298s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0867 Mean episode length: 1340.5 Actor loss: -9.4770 Critic loss: +0.0093 World model loss: +0.5554 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4927 Episode_Reward/track_angular_velocity: +50.5802 Episode_Reward/upright: +25.6832 Episode_Reward/leg_posture: -2.5296 Episode_Reward/standstill: -6.9720 Episode_Reward/action_rate_l2: -0.5692 Episode_Reward/joint_torques_l2: -0.1426 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0711 Episode_Reward/rolling_leg_motion: -0.5528 Episode_Reward/obstacle_progress: +3.8410 Episode_Reward/wheel_body_intrusion: -0.2736 Episode_Reward/pair_body_intrusion: -0.2598 Curriculum/terrain_level_mean: +4.2642 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4394 Episode_Termination/fell_over: 0.0379 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== ============================================================================== Learning iteration 45200/50000 Total steps: 185,139,200 Steps per second: 13,752 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1317.2 Actor loss: -9.5082 Critic loss: +0.0067 World model loss: +0.5982 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7029 Episode_Reward/track_angular_velocity: +49.7255 Episode_Reward/upright: +25.2180 Episode_Reward/leg_posture: -2.5399 Episode_Reward/standstill: -6.2739 Episode_Reward/action_rate_l2: -0.5520 Episode_Reward/joint_torques_l2: -0.1410 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0386 Episode_Reward/unnecessary_wheel_lift: -0.0910 Episode_Reward/rolling_leg_motion: -0.5706 Episode_Reward/obstacle_progress: +3.6550 Episode_Reward/wheel_body_intrusion: -0.2651 Episode_Reward/pair_body_intrusion: -0.2566 Curriculum/terrain_level_mean: +4.2644 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.4372 Episode_Termination/fell_over: 0.0383 Episode_Termination/course_exit: 0.4809 Episode_Termination/illegal_nonwheel_contact: 0.0546 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1039] ============================================================================== Learning iteration 45250/50000 Total steps: 185,344,000 Steps per second: 13,754 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1373.4 Actor loss: -9.4982 Critic loss: +0.0061 World model loss: +0.5880 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8300 Episode_Reward/track_angular_velocity: +51.8590 Episode_Reward/upright: +26.3051 Episode_Reward/leg_posture: -2.6418 Episode_Reward/standstill: -6.5418 Episode_Reward/action_rate_l2: -0.5933 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0054 Episode_Reward/unnecessary_wheel_lift: -0.1114 Episode_Reward/rolling_leg_motion: -0.6019 Episode_Reward/obstacle_progress: +3.9289 Episode_Reward/wheel_body_intrusion: -0.2807 Episode_Reward/pair_body_intrusion: -0.2851 Curriculum/terrain_level_mean: +4.2664 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4750 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.5125 Episode_Termination/illegal_nonwheel_contact: 0.0125 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[994] RECOVERY reset 1 invalid environments; first IDs=[1390] RECOVERY reset 1 invalid environments; first IDs=[1019] ============================================================================== Learning iteration 45300/50000 Total steps: 185,548,800 Steps per second: 13,756 Iteration time: 0.298s Collection time: 0.077s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1336.5 Actor loss: -9.5346 Critic loss: +0.0077 World model loss: +0.5712 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3288 Episode_Reward/track_angular_velocity: +50.4465 Episode_Reward/upright: +25.5872 Episode_Reward/leg_posture: -2.5305 Episode_Reward/standstill: -6.0909 Episode_Reward/action_rate_l2: -0.5719 Episode_Reward/joint_torques_l2: -0.1420 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.1008 Episode_Reward/rolling_leg_motion: -0.5746 Episode_Reward/obstacle_progress: +3.7865 Episode_Reward/wheel_body_intrusion: -0.2679 Episode_Reward/pair_body_intrusion: -0.2981 Curriculum/terrain_level_mean: +4.2659 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4726 Episode_Termination/fell_over: 0.0274 Episode_Termination/course_exit: 0.4452 Episode_Termination/illegal_nonwheel_contact: 0.0616 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[978] ============================================================================== Learning iteration 45350/50000 Total steps: 185,753,600 Steps per second: 13,758 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1381.3 Actor loss: -9.5234 Critic loss: +0.0036 World model loss: +0.5658 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9930 Episode_Reward/track_angular_velocity: +52.2034 Episode_Reward/upright: +26.4399 Episode_Reward/leg_posture: -2.6358 Episode_Reward/standstill: -6.3171 Episode_Reward/action_rate_l2: -0.5855 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0270 Episode_Reward/unnecessary_wheel_lift: -0.0781 Episode_Reward/rolling_leg_motion: -0.5893 Episode_Reward/obstacle_progress: +3.9144 Episode_Reward/wheel_body_intrusion: -0.2561 Episode_Reward/pair_body_intrusion: -0.2926 Curriculum/terrain_level_mean: +4.2659 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.4898 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4422 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2338] ============================================================================== Learning iteration 45400/50000 Total steps: 185,958,400 Steps per second: 13,760 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1351.2 Actor loss: -9.5134 Critic loss: +0.0060 World model loss: +0.5526 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9536 Episode_Reward/track_angular_velocity: +51.0347 Episode_Reward/upright: +25.8257 Episode_Reward/leg_posture: -2.5816 Episode_Reward/standstill: -6.5251 Episode_Reward/action_rate_l2: -0.5742 Episode_Reward/joint_torques_l2: -0.1449 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5705 Episode_Reward/obstacle_progress: +3.8442 Episode_Reward/wheel_body_intrusion: -0.2522 Episode_Reward/pair_body_intrusion: -0.2497 Curriculum/terrain_level_mean: +4.2661 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4679 Episode_Termination/fell_over: 0.0449 Episode_Termination/course_exit: 0.4744 Episode_Termination/illegal_nonwheel_contact: 0.0321 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2177] ============================================================================== Learning iteration 45450/50000 Total steps: 186,163,200 Steps per second: 13,761 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1338.6 Actor loss: -9.4907 Critic loss: +0.0058 World model loss: +0.5676 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3062 Episode_Reward/track_angular_velocity: +50.5331 Episode_Reward/upright: +25.6218 Episode_Reward/leg_posture: -2.5307 Episode_Reward/standstill: -6.3549 Episode_Reward/action_rate_l2: -0.5711 Episode_Reward/joint_torques_l2: -0.1433 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0314 Episode_Reward/unnecessary_wheel_lift: -0.0831 Episode_Reward/rolling_leg_motion: -0.5735 Episode_Reward/obstacle_progress: +3.7209 Episode_Reward/wheel_body_intrusion: -0.2487 Episode_Reward/pair_body_intrusion: -0.2668 Curriculum/terrain_level_mean: +4.2666 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4621 Episode_Termination/fell_over: 0.0303 Episode_Termination/course_exit: 0.4470 Episode_Termination/illegal_nonwheel_contact: 0.0758 ============================================================================== ============================================================================== Learning iteration 45500/50000 Total steps: 186,368,000 Steps per second: 13,763 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0871 Mean episode length: 1413.7 Actor loss: -9.4811 Critic loss: +0.0050 World model loss: +0.5727 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0054 Episode_Reward/track_angular_velocity: +53.3908 Episode_Reward/upright: +27.1011 Episode_Reward/leg_posture: -2.6253 Episode_Reward/standstill: -6.6267 Episode_Reward/action_rate_l2: -0.6104 Episode_Reward/joint_torques_l2: -0.1496 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0249 Episode_Reward/unnecessary_wheel_lift: -0.0951 Episode_Reward/rolling_leg_motion: -0.6150 Episode_Reward/obstacle_progress: +3.9319 Episode_Reward/wheel_body_intrusion: -0.2611 Episode_Reward/pair_body_intrusion: -0.2717 Curriculum/terrain_level_mean: +4.2673 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.5349 Episode_Termination/fell_over: 0.0349 Episode_Termination/course_exit: 0.4128 Episode_Termination/illegal_nonwheel_contact: 0.0233 ============================================================================== ============================================================================== Learning iteration 45550/50000 Total steps: 186,572,800 Steps per second: 13,765 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1367.2 Actor loss: -9.5269 Critic loss: +0.0050 World model loss: +0.5713 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4246 Episode_Reward/track_angular_velocity: +51.6193 Episode_Reward/upright: +26.2116 Episode_Reward/leg_posture: -2.5726 Episode_Reward/standstill: -6.0293 Episode_Reward/action_rate_l2: -0.5822 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0080 Episode_Reward/unnecessary_wheel_lift: -0.0850 Episode_Reward/rolling_leg_motion: -0.5882 Episode_Reward/obstacle_progress: +3.7372 Episode_Reward/wheel_body_intrusion: -0.2518 Episode_Reward/pair_body_intrusion: -0.2537 Curriculum/terrain_level_mean: +4.2708 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4632 Episode_Termination/fell_over: 0.0147 Episode_Termination/course_exit: 0.5147 Episode_Termination/illegal_nonwheel_contact: 0.0074 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[965] ============================================================================== Learning iteration 45600/50000 Total steps: 186,777,600 Steps per second: 13,767 Iteration time: 0.298s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1322.7 Actor loss: -9.5088 Critic loss: +0.0057 World model loss: +0.5979 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.7654 Episode_Reward/track_angular_velocity: +49.9511 Episode_Reward/upright: +25.3386 Episode_Reward/leg_posture: -2.4989 Episode_Reward/standstill: -5.8967 Episode_Reward/action_rate_l2: -0.5573 Episode_Reward/joint_torques_l2: -0.1411 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0207 Episode_Reward/unnecessary_wheel_lift: -0.0813 Episode_Reward/rolling_leg_motion: -0.5628 Episode_Reward/obstacle_progress: +3.6566 Episode_Reward/wheel_body_intrusion: -0.2311 Episode_Reward/pair_body_intrusion: -0.2507 Curriculum/terrain_level_mean: +4.2705 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4370 Episode_Termination/fell_over: 0.0336 Episode_Termination/course_exit: 0.4958 Episode_Termination/illegal_nonwheel_contact: 0.0336 ============================================================================== ============================================================================== Learning iteration 45650/50000 Total steps: 186,982,400 Steps per second: 13,769 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1373.0 Actor loss: -9.4853 Critic loss: +0.0051 World model loss: +0.6115 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5703 Episode_Reward/track_angular_velocity: +51.8824 Episode_Reward/upright: +26.3049 Episode_Reward/leg_posture: -2.5483 Episode_Reward/standstill: -6.3196 Episode_Reward/action_rate_l2: -0.5876 Episode_Reward/joint_torques_l2: -0.1472 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0295 Episode_Reward/unnecessary_wheel_lift: -0.0710 Episode_Reward/rolling_leg_motion: -0.5686 Episode_Reward/obstacle_progress: +3.9003 Episode_Reward/wheel_body_intrusion: -0.2501 Episode_Reward/pair_body_intrusion: -0.2815 Curriculum/terrain_level_mean: +4.2710 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4892 Episode_Termination/fell_over: 0.0360 Episode_Termination/course_exit: 0.4317 Episode_Termination/illegal_nonwheel_contact: 0.0504 ============================================================================== ============================================================================== Learning iteration 45700/50000 Total steps: 187,187,200 Steps per second: 13,770 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1371.4 Actor loss: -9.5062 Critic loss: +0.0085 World model loss: +0.5321 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5902 Episode_Reward/track_angular_velocity: +51.8155 Episode_Reward/upright: +26.2829 Episode_Reward/leg_posture: -2.5628 Episode_Reward/standstill: -6.1540 Episode_Reward/action_rate_l2: -0.5903 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0832 Episode_Reward/rolling_leg_motion: -0.5744 Episode_Reward/obstacle_progress: +3.8312 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.3117 Curriculum/terrain_level_mean: +4.2732 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1348 Episode_Termination/time_out: 0.4654 Episode_Termination/fell_over: 0.0126 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0189 ============================================================================== ============================================================================== Learning iteration 45750/50000 Total steps: 187,392,000 Steps per second: 13,772 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1323.9 Actor loss: -9.5299 Critic loss: +0.0051 World model loss: +0.5453 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9703 Episode_Reward/track_angular_velocity: +50.0017 Episode_Reward/upright: +25.3534 Episode_Reward/leg_posture: -2.5244 Episode_Reward/standstill: -6.1505 Episode_Reward/action_rate_l2: -0.5567 Episode_Reward/joint_torques_l2: -0.1395 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0841 Episode_Reward/rolling_leg_motion: -0.5686 Episode_Reward/obstacle_progress: +3.6175 Episode_Reward/wheel_body_intrusion: -0.2661 Episode_Reward/pair_body_intrusion: -0.2806 Curriculum/terrain_level_mean: +4.2766 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4528 Episode_Termination/fell_over: 0.0252 Episode_Termination/course_exit: 0.5094 Episode_Termination/illegal_nonwheel_contact: 0.0189 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[667] ============================================================================== Learning iteration 45800/50000 Total steps: 187,596,800 Steps per second: 13,774 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0867 Mean episode length: 1303.7 Actor loss: -9.5284 Critic loss: +0.0048 World model loss: +0.5507 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.1746 Episode_Reward/track_angular_velocity: +49.1866 Episode_Reward/upright: +24.9613 Episode_Reward/leg_posture: -2.4948 Episode_Reward/standstill: -6.4484 Episode_Reward/action_rate_l2: -0.5542 Episode_Reward/joint_torques_l2: -0.1382 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0280 Episode_Reward/unnecessary_wheel_lift: -0.0885 Episode_Reward/rolling_leg_motion: -0.5630 Episode_Reward/obstacle_progress: +3.6015 Episode_Reward/wheel_body_intrusion: -0.2599 Episode_Reward/pair_body_intrusion: -0.2805 Curriculum/terrain_level_mean: +4.2776 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4423 Episode_Termination/fell_over: 0.0577 Episode_Termination/course_exit: 0.4872 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== ============================================================================== Learning iteration 45850/50000 Total steps: 187,801,600 Steps per second: 13,776 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1374.7 Actor loss: -9.4718 Critic loss: +0.0051 World model loss: +0.5849 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7506 Episode_Reward/track_angular_velocity: +51.9037 Episode_Reward/upright: +26.3613 Episode_Reward/leg_posture: -2.6343 Episode_Reward/standstill: -6.7069 Episode_Reward/action_rate_l2: -0.5805 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0871 Episode_Reward/rolling_leg_motion: -0.5910 Episode_Reward/obstacle_progress: +3.7977 Episode_Reward/wheel_body_intrusion: -0.2809 Episode_Reward/pair_body_intrusion: -0.2698 Curriculum/terrain_level_mean: +4.2793 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.5153 Episode_Termination/fell_over: 0.0000 Episode_Termination/course_exit: 0.4540 Episode_Termination/illegal_nonwheel_contact: 0.0307 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1826] ============================================================================== Learning iteration 45900/50000 Total steps: 188,006,400 Steps per second: 13,778 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1385.8 Actor loss: -9.5118 Critic loss: +0.0061 World model loss: +0.5656 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2052 Episode_Reward/track_angular_velocity: +52.3744 Episode_Reward/upright: +26.5558 Episode_Reward/leg_posture: -2.6412 Episode_Reward/standstill: -6.2813 Episode_Reward/action_rate_l2: -0.5843 Episode_Reward/joint_torques_l2: -0.1471 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0804 Episode_Reward/rolling_leg_motion: -0.5834 Episode_Reward/obstacle_progress: +3.9135 Episode_Reward/wheel_body_intrusion: -0.2618 Episode_Reward/pair_body_intrusion: -0.2825 Curriculum/terrain_level_mean: +4.2805 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5364 Episode_Termination/fell_over: 0.0331 Episode_Termination/course_exit: 0.4238 Episode_Termination/illegal_nonwheel_contact: 0.0132 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_045900.mp4 RECOVERY reset 1 invalid environments; first IDs=[1325] ============================================================================== Learning iteration 45950/50000 Total steps: 188,211,200 Steps per second: 13,774 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1402.3 Actor loss: -9.5131 Critic loss: +0.0052 World model loss: +0.5674 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7801 Episode_Reward/track_angular_velocity: +53.0240 Episode_Reward/upright: +26.8672 Episode_Reward/leg_posture: -2.6968 Episode_Reward/standstill: -6.8541 Episode_Reward/action_rate_l2: -0.5927 Episode_Reward/joint_torques_l2: -0.1484 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0219 Episode_Reward/unnecessary_wheel_lift: -0.0920 Episode_Reward/rolling_leg_motion: -0.6096 Episode_Reward/obstacle_progress: +3.8988 Episode_Reward/wheel_body_intrusion: -0.2871 Episode_Reward/pair_body_intrusion: -0.2831 Curriculum/terrain_level_mean: +4.2815 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1336 Episode_Termination/time_out: 0.4880 Episode_Termination/fell_over: 0.0422 Episode_Termination/course_exit: 0.4639 Episode_Termination/illegal_nonwheel_contact: 0.0181 ============================================================================== ============================================================================== Learning iteration 46000/50000 Total steps: 188,416,000 Steps per second: 13,776 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0872 Mean episode length: 1340.0 Actor loss: -9.4993 Critic loss: +0.0050 World model loss: +0.5761 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5823 Episode_Reward/track_angular_velocity: +50.7119 Episode_Reward/upright: +25.6533 Episode_Reward/leg_posture: -2.6371 Episode_Reward/standstill: -6.2893 Episode_Reward/action_rate_l2: -0.5624 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0826 Episode_Reward/rolling_leg_motion: -0.5770 Episode_Reward/obstacle_progress: +3.7070 Episode_Reward/wheel_body_intrusion: -0.2760 Episode_Reward/pair_body_intrusion: -0.2368 Curriculum/terrain_level_mean: +4.2830 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1326 Episode_Termination/time_out: 0.4730 Episode_Termination/fell_over: 0.0203 Episode_Termination/course_exit: 0.4730 Episode_Termination/illegal_nonwheel_contact: 0.0405 ============================================================================== ============================================================================== Learning iteration 46050/50000 Total steps: 188,620,800 Steps per second: 13,778 Iteration time: 0.297s Collection time: 0.075s Learning time: 0.183s Mean reward: +0.0872 Mean episode length: 1349.8 Actor loss: -9.5376 Critic loss: +0.0039 World model loss: +0.5431 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9351 Episode_Reward/track_angular_velocity: +51.0255 Episode_Reward/upright: +25.8331 Episode_Reward/leg_posture: -2.6246 Episode_Reward/standstill: -6.3120 Episode_Reward/action_rate_l2: -0.5635 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0079 Episode_Reward/unnecessary_wheel_lift: -0.0804 Episode_Reward/rolling_leg_motion: -0.5808 Episode_Reward/obstacle_progress: +3.7881 Episode_Reward/wheel_body_intrusion: -0.2596 Episode_Reward/pair_body_intrusion: -0.2750 Curriculum/terrain_level_mean: +4.2842 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4651 Episode_Termination/fell_over: 0.0233 Episode_Termination/course_exit: 0.5116 Episode_Termination/illegal_nonwheel_contact: 0.0155 ============================================================================== ============================================================================== Learning iteration 46100/50000 Total steps: 188,825,600 Steps per second: 13,779 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0868 Mean episode length: 1335.5 Actor loss: -9.5157 Critic loss: +0.0032 World model loss: +0.5538 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3835 Episode_Reward/track_angular_velocity: +50.3493 Episode_Reward/upright: +25.5107 Episode_Reward/leg_posture: -2.5881 Episode_Reward/standstill: -6.5257 Episode_Reward/action_rate_l2: -0.5603 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0313 Episode_Reward/unnecessary_wheel_lift: -0.1043 Episode_Reward/rolling_leg_motion: -0.5708 Episode_Reward/obstacle_progress: +3.7196 Episode_Reward/wheel_body_intrusion: -0.2722 Episode_Reward/pair_body_intrusion: -0.2725 Curriculum/terrain_level_mean: +4.2847 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4586 Episode_Termination/fell_over: 0.0573 Episode_Termination/course_exit: 0.4650 Episode_Termination/illegal_nonwheel_contact: 0.0573 ============================================================================== ============================================================================== Learning iteration 46150/50000 Total steps: 189,030,400 Steps per second: 13,781 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0869 Mean episode length: 1337.8 Actor loss: -9.5182 Critic loss: +0.0036 World model loss: +0.5598 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4396 Episode_Reward/track_angular_velocity: +50.4710 Episode_Reward/upright: +25.5824 Episode_Reward/leg_posture: -2.6534 Episode_Reward/standstill: -6.3042 Episode_Reward/action_rate_l2: -0.5561 Episode_Reward/joint_torques_l2: -0.1431 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0235 Episode_Reward/unnecessary_wheel_lift: -0.0865 Episode_Reward/rolling_leg_motion: -0.5717 Episode_Reward/obstacle_progress: +3.6279 Episode_Reward/wheel_body_intrusion: -0.2680 Episode_Reward/pair_body_intrusion: -0.2592 Curriculum/terrain_level_mean: +4.2866 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0930 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5152 Episode_Termination/fell_over: 0.0182 Episode_Termination/course_exit: 0.4303 Episode_Termination/illegal_nonwheel_contact: 0.0485 ============================================================================== ============================================================================== Learning iteration 46200/50000 Total steps: 189,235,200 Steps per second: 13,783 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1392.8 Actor loss: -9.5088 Critic loss: +0.0034 World model loss: +0.5637 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2613 Episode_Reward/track_angular_velocity: +52.4508 Episode_Reward/upright: +26.6021 Episode_Reward/leg_posture: -2.7232 Episode_Reward/standstill: -6.1580 Episode_Reward/action_rate_l2: -0.5849 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0299 Episode_Reward/unnecessary_wheel_lift: -0.0775 Episode_Reward/rolling_leg_motion: -0.5865 Episode_Reward/obstacle_progress: +3.8081 Episode_Reward/wheel_body_intrusion: -0.2848 Episode_Reward/pair_body_intrusion: -0.2803 Curriculum/terrain_level_mean: +4.2871 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0943 Metrics/base_velocity/error_vel_yaw: 0.1382 Episode_Termination/time_out: 0.5033 Episode_Termination/fell_over: 0.0523 Episode_Termination/course_exit: 0.4248 Episode_Termination/illegal_nonwheel_contact: 0.0392 ============================================================================== ============================================================================== Learning iteration 46250/50000 Total steps: 189,440,000 Steps per second: 13,785 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0870 Mean episode length: 1353.1 Actor loss: -9.4802 Critic loss: +0.0072 World model loss: +0.5768 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7389 Episode_Reward/track_angular_velocity: +51.0051 Episode_Reward/upright: +25.8587 Episode_Reward/leg_posture: -2.6352 Episode_Reward/standstill: -5.9891 Episode_Reward/action_rate_l2: -0.5748 Episode_Reward/joint_torques_l2: -0.1443 Episode_Reward/dof_pos_limits: -0.0012 Episode_Reward/terminated: -0.0344 Episode_Reward/unnecessary_wheel_lift: -0.0940 Episode_Reward/rolling_leg_motion: -0.5792 Episode_Reward/obstacle_progress: +3.7173 Episode_Reward/wheel_body_intrusion: -0.2569 Episode_Reward/pair_body_intrusion: -0.2662 Curriculum/terrain_level_mean: +4.2866 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0954 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4620 Episode_Termination/fell_over: 0.0506 Episode_Termination/course_exit: 0.4557 Episode_Termination/illegal_nonwheel_contact: 0.0443 ============================================================================== ============================================================================== Learning iteration 46300/50000 Total steps: 189,644,800 Steps per second: 13,787 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1344.2 Actor loss: -9.4872 Critic loss: +0.0060 World model loss: +0.5407 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5095 Episode_Reward/track_angular_velocity: +50.7049 Episode_Reward/upright: +25.7254 Episode_Reward/leg_posture: -2.5731 Episode_Reward/standstill: -5.7003 Episode_Reward/action_rate_l2: -0.5678 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0814 Episode_Reward/rolling_leg_motion: -0.5692 Episode_Reward/obstacle_progress: +3.7431 Episode_Reward/wheel_body_intrusion: -0.2442 Episode_Reward/pair_body_intrusion: -0.2683 Curriculum/terrain_level_mean: +4.2874 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4085 Episode_Termination/fell_over: 0.0211 Episode_Termination/course_exit: 0.5423 Episode_Termination/illegal_nonwheel_contact: 0.0352 ============================================================================== ============================================================================== Learning iteration 46350/50000 Total steps: 189,849,600 Steps per second: 13,788 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.193s Mean reward: +0.0877 Mean episode length: 1391.3 Actor loss: -9.5158 Critic loss: +0.0036 World model loss: +0.5814 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.2986 Episode_Reward/track_angular_velocity: +52.5095 Episode_Reward/upright: +26.6727 Episode_Reward/leg_posture: -2.6301 Episode_Reward/standstill: -5.7143 Episode_Reward/action_rate_l2: -0.5896 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0109 Episode_Reward/unnecessary_wheel_lift: -0.0747 Episode_Reward/rolling_leg_motion: -0.5863 Episode_Reward/obstacle_progress: +3.8821 Episode_Reward/wheel_body_intrusion: -0.2643 Episode_Reward/pair_body_intrusion: -0.2793 Curriculum/terrain_level_mean: +4.2891 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.4759 Episode_Termination/fell_over: 0.0120 Episode_Termination/course_exit: 0.5000 Episode_Termination/illegal_nonwheel_contact: 0.0181 ============================================================================== ============================================================================== Learning iteration 46400/50000 Total steps: 190,054,400 Steps per second: 13,790 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1405.5 Actor loss: -9.5117 Critic loss: +0.0072 World model loss: +0.5913 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8029 Episode_Reward/track_angular_velocity: +53.0379 Episode_Reward/upright: +26.9328 Episode_Reward/leg_posture: -2.6508 Episode_Reward/standstill: -5.8822 Episode_Reward/action_rate_l2: -0.5976 Episode_Reward/joint_torques_l2: -0.1493 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0055 Episode_Reward/unnecessary_wheel_lift: -0.0750 Episode_Reward/rolling_leg_motion: -0.6024 Episode_Reward/obstacle_progress: +3.9302 Episode_Reward/wheel_body_intrusion: -0.2825 Episode_Reward/pair_body_intrusion: -0.3144 Curriculum/terrain_level_mean: +4.2915 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1360 Episode_Termination/time_out: 0.5195 Episode_Termination/fell_over: 0.0065 Episode_Termination/course_exit: 0.4675 Episode_Termination/illegal_nonwheel_contact: 0.0065 ============================================================================== ============================================================================== Learning iteration 46450/50000 Total steps: 190,259,200 Steps per second: 13,792 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1336.4 Actor loss: -9.5166 Critic loss: +0.0048 World model loss: +0.5701 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3735 Episode_Reward/track_angular_velocity: +50.4889 Episode_Reward/upright: +25.5899 Episode_Reward/leg_posture: -2.5549 Episode_Reward/standstill: -5.8528 Episode_Reward/action_rate_l2: -0.5598 Episode_Reward/joint_torques_l2: -0.1417 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0762 Episode_Reward/rolling_leg_motion: -0.5656 Episode_Reward/obstacle_progress: +3.7616 Episode_Reward/wheel_body_intrusion: -0.2609 Episode_Reward/pair_body_intrusion: -0.2663 Curriculum/terrain_level_mean: +4.2932 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4340 Episode_Termination/fell_over: 0.0126 Episode_Termination/course_exit: 0.5283 Episode_Termination/illegal_nonwheel_contact: 0.0252 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1890] ============================================================================== Learning iteration 46500/50000 Total steps: 190,464,000 Steps per second: 13,794 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1354.0 Actor loss: -9.5006 Critic loss: +0.0043 World model loss: +0.5286 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9767 Episode_Reward/track_angular_velocity: +51.1345 Episode_Reward/upright: +25.9130 Episode_Reward/leg_posture: -2.6108 Episode_Reward/standstill: -6.0188 Episode_Reward/action_rate_l2: -0.5677 Episode_Reward/joint_torques_l2: -0.1438 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0344 Episode_Reward/unnecessary_wheel_lift: -0.0751 Episode_Reward/rolling_leg_motion: -0.5676 Episode_Reward/obstacle_progress: +3.8321 Episode_Reward/wheel_body_intrusion: -0.2671 Episode_Reward/pair_body_intrusion: -0.2708 Curriculum/terrain_level_mean: +4.2922 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4571 Episode_Termination/fell_over: 0.0571 Episode_Termination/course_exit: 0.4500 Episode_Termination/illegal_nonwheel_contact: 0.0429 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 46550/50000 Total steps: 190,668,800 Steps per second: 13,795 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0873 Mean episode length: 1374.3 Actor loss: -9.4731 Critic loss: +0.0122 World model loss: +0.5593 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7335 Episode_Reward/track_angular_velocity: +51.9366 Episode_Reward/upright: +26.2967 Episode_Reward/leg_posture: -2.6576 Episode_Reward/standstill: -6.2704 Episode_Reward/action_rate_l2: -0.5798 Episode_Reward/joint_torques_l2: -0.1463 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0322 Episode_Reward/unnecessary_wheel_lift: -0.0726 Episode_Reward/rolling_leg_motion: -0.5823 Episode_Reward/obstacle_progress: +3.8869 Episode_Reward/wheel_body_intrusion: -0.2791 Episode_Reward/pair_body_intrusion: -0.2873 Curriculum/terrain_level_mean: +4.2922 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.5132 Episode_Termination/fell_over: 0.0395 Episode_Termination/course_exit: 0.4079 Episode_Termination/illegal_nonwheel_contact: 0.0461 ============================================================================== ============================================================================== Learning iteration 46600/50000 Total steps: 190,873,600 Steps per second: 13,797 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1356.0 Actor loss: -9.4778 Critic loss: +0.0048 World model loss: +0.5174 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0545 Episode_Reward/track_angular_velocity: +51.2142 Episode_Reward/upright: +25.9629 Episode_Reward/leg_posture: -2.5911 Episode_Reward/standstill: -6.0202 Episode_Reward/action_rate_l2: -0.5717 Episode_Reward/joint_torques_l2: -0.1430 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0685 Episode_Reward/rolling_leg_motion: -0.5727 Episode_Reward/obstacle_progress: +3.7760 Episode_Reward/wheel_body_intrusion: -0.2569 Episode_Reward/pair_body_intrusion: -0.2745 Curriculum/terrain_level_mean: +4.2944 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0268 Episode_Termination/course_exit: 0.4631 Episode_Termination/illegal_nonwheel_contact: 0.0134 ============================================================================== ============================================================================== Learning iteration 46650/50000 Total steps: 191,078,400 Steps per second: 13,799 Iteration time: 0.297s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1347.3 Actor loss: -9.5044 Critic loss: +0.0039 World model loss: +0.5591 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7709 Episode_Reward/track_angular_velocity: +50.8520 Episode_Reward/upright: +25.8204 Episode_Reward/leg_posture: -2.5923 Episode_Reward/standstill: -6.0715 Episode_Reward/action_rate_l2: -0.5608 Episode_Reward/joint_torques_l2: -0.1425 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0079 Episode_Reward/unnecessary_wheel_lift: -0.0818 Episode_Reward/rolling_leg_motion: -0.5774 Episode_Reward/obstacle_progress: +3.7742 Episode_Reward/wheel_body_intrusion: -0.2723 Episode_Reward/pair_body_intrusion: -0.2862 Curriculum/terrain_level_mean: +4.2969 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4366 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.5423 Episode_Termination/illegal_nonwheel_contact: 0.0211 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[678] RECOVERY reset 1 invalid environments; first IDs=[916] ============================================================================== Learning iteration 46700/50000 Total steps: 191,283,200 Steps per second: 13,801 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1322.7 Actor loss: -9.4978 Critic loss: +0.0066 World model loss: +0.5445 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.8378 Episode_Reward/track_angular_velocity: +49.9731 Episode_Reward/upright: +25.3517 Episode_Reward/leg_posture: -2.5675 Episode_Reward/standstill: -6.0992 Episode_Reward/action_rate_l2: -0.5620 Episode_Reward/joint_torques_l2: -0.1415 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0155 Episode_Reward/unnecessary_wheel_lift: -0.0758 Episode_Reward/rolling_leg_motion: -0.5689 Episode_Reward/obstacle_progress: +3.5747 Episode_Reward/wheel_body_intrusion: -0.2753 Episode_Reward/pair_body_intrusion: -0.2892 Curriculum/terrain_level_mean: +4.2974 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4539 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.5066 Episode_Termination/illegal_nonwheel_contact: 0.0263 ============================================================================== ============================================================================== Learning iteration 46750/50000 Total steps: 191,488,000 Steps per second: 13,803 Iteration time: 0.297s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1413.0 Actor loss: -9.5053 Critic loss: +0.0036 World model loss: +0.5686 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.8822 Episode_Reward/track_angular_velocity: +53.4324 Episode_Reward/upright: +27.1152 Episode_Reward/leg_posture: -2.6885 Episode_Reward/standstill: -6.7353 Episode_Reward/action_rate_l2: -0.5890 Episode_Reward/joint_torques_l2: -0.1493 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0138 Episode_Reward/unnecessary_wheel_lift: -0.0924 Episode_Reward/rolling_leg_motion: -0.6016 Episode_Reward/obstacle_progress: +3.7746 Episode_Reward/wheel_body_intrusion: -0.2504 Episode_Reward/pair_body_intrusion: -0.2724 Curriculum/terrain_level_mean: +4.2986 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0952 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.5177 Episode_Termination/fell_over: 0.0071 Episode_Termination/course_exit: 0.4468 Episode_Termination/illegal_nonwheel_contact: 0.0355 ============================================================================== ============================================================================== Learning iteration 46800/50000 Total steps: 191,692,800 Steps per second: 13,804 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1356.5 Actor loss: -9.5136 Critic loss: +0.0055 World model loss: +0.5486 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1564 Episode_Reward/track_angular_velocity: +51.2870 Episode_Reward/upright: +26.0550 Episode_Reward/leg_posture: -2.5573 Episode_Reward/standstill: -6.1751 Episode_Reward/action_rate_l2: -0.5577 Episode_Reward/joint_torques_l2: -0.1413 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.0900 Episode_Reward/rolling_leg_motion: -0.5835 Episode_Reward/obstacle_progress: +3.6886 Episode_Reward/wheel_body_intrusion: -0.2443 Episode_Reward/pair_body_intrusion: -0.3049 Curriculum/terrain_level_mean: +4.2983 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1337 Episode_Termination/time_out: 0.4422 Episode_Termination/fell_over: 0.0476 Episode_Termination/course_exit: 0.4966 Episode_Termination/illegal_nonwheel_contact: 0.0340 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2777] ============================================================================== Learning iteration 46850/50000 Total steps: 191,897,600 Steps per second: 13,806 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1274.8 Actor loss: -9.4291 Critic loss: +0.0100 World model loss: +0.5554 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.1930 Episode_Reward/track_angular_velocity: +48.1577 Episode_Reward/upright: +24.4585 Episode_Reward/leg_posture: -2.4230 Episode_Reward/standstill: -6.0319 Episode_Reward/action_rate_l2: -0.5196 Episode_Reward/joint_torques_l2: -0.1335 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0124 Episode_Reward/unnecessary_wheel_lift: -0.0632 Episode_Reward/rolling_leg_motion: -0.5270 Episode_Reward/obstacle_progress: +3.4965 Episode_Reward/wheel_body_intrusion: -0.2440 Episode_Reward/pair_body_intrusion: -0.2915 Curriculum/terrain_level_mean: +4.3003 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0929 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.3901 Episode_Termination/fell_over: 0.0142 Episode_Termination/course_exit: 0.5745 Episode_Termination/illegal_nonwheel_contact: 0.0213 ============================================================================== ============================================================================== Learning iteration 46900/50000 Total steps: 192,102,400 Steps per second: 13,808 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1323.5 Actor loss: -9.4951 Critic loss: +0.0039 World model loss: +0.5760 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.9656 Episode_Reward/track_angular_velocity: +49.9897 Episode_Reward/upright: +25.3304 Episode_Reward/leg_posture: -2.5292 Episode_Reward/standstill: -6.3017 Episode_Reward/action_rate_l2: -0.5445 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0181 Episode_Reward/unnecessary_wheel_lift: -0.0766 Episode_Reward/rolling_leg_motion: -0.5569 Episode_Reward/obstacle_progress: +3.6979 Episode_Reward/wheel_body_intrusion: -0.2584 Episode_Reward/pair_body_intrusion: -0.2624 Curriculum/terrain_level_mean: +4.3013 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4599 Episode_Termination/fell_over: 0.0073 Episode_Termination/course_exit: 0.4891 Episode_Termination/illegal_nonwheel_contact: 0.0511 ============================================================================== ============================================================================== Learning iteration 46950/50000 Total steps: 192,307,200 Steps per second: 13,810 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1370.6 Actor loss: -9.5211 Critic loss: +0.0041 World model loss: +0.5323 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7477 Episode_Reward/track_angular_velocity: +51.7720 Episode_Reward/upright: +26.2534 Episode_Reward/leg_posture: -2.6507 Episode_Reward/standstill: -6.2340 Episode_Reward/action_rate_l2: -0.5595 Episode_Reward/joint_torques_l2: -0.1459 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0816 Episode_Reward/rolling_leg_motion: -0.5877 Episode_Reward/obstacle_progress: +3.8603 Episode_Reward/wheel_body_intrusion: -0.2847 Episode_Reward/pair_body_intrusion: -0.2808 Curriculum/terrain_level_mean: +4.3020 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5305 Episode_Termination/fell_over: 0.0183 Episode_Termination/course_exit: 0.4329 Episode_Termination/illegal_nonwheel_contact: 0.0244 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[887] ============================================================================== Learning iteration 47000/50000 Total steps: 192,512,000 Steps per second: 13,811 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0874 Mean episode length: 1380.9 Actor loss: -9.5145 Critic loss: +0.0059 World model loss: +0.5433 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1300 Episode_Reward/track_angular_velocity: +52.1824 Episode_Reward/upright: +26.4660 Episode_Reward/leg_posture: -2.6243 Episode_Reward/standstill: -6.4267 Episode_Reward/action_rate_l2: -0.5604 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0783 Episode_Reward/rolling_leg_motion: -0.5882 Episode_Reward/obstacle_progress: +3.9204 Episode_Reward/wheel_body_intrusion: -0.2661 Episode_Reward/pair_body_intrusion: -0.3102 Curriculum/terrain_level_mean: +4.3032 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0149 Episode_Termination/course_exit: 0.4701 Episode_Termination/illegal_nonwheel_contact: 0.0149 ============================================================================== ============================================================================== Learning iteration 47050/50000 Total steps: 192,716,800 Steps per second: 13,813 Iteration time: 0.297s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1371.0 Actor loss: -9.5346 Critic loss: +0.0037 World model loss: +0.5462 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6929 Episode_Reward/track_angular_velocity: +51.8011 Episode_Reward/upright: +26.2863 Episode_Reward/leg_posture: -2.5595 Episode_Reward/standstill: -6.0673 Episode_Reward/action_rate_l2: -0.5684 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0241 Episode_Reward/unnecessary_wheel_lift: -0.0793 Episode_Reward/rolling_leg_motion: -0.5809 Episode_Reward/obstacle_progress: +3.9292 Episode_Reward/wheel_body_intrusion: -0.2374 Episode_Reward/pair_body_intrusion: -0.3177 Curriculum/terrain_level_mean: +4.3037 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4695 Episode_Termination/fell_over: 0.0366 Episode_Termination/course_exit: 0.4756 Episode_Termination/illegal_nonwheel_contact: 0.0305 ============================================================================== ============================================================================== Learning iteration 47100/50000 Total steps: 192,921,600 Steps per second: 13,815 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1373.3 Actor loss: -9.5043 Critic loss: +0.0045 World model loss: +0.5803 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.7797 Episode_Reward/track_angular_velocity: +51.9204 Episode_Reward/upright: +26.3303 Episode_Reward/leg_posture: -2.5833 Episode_Reward/standstill: -6.1792 Episode_Reward/action_rate_l2: -0.5810 Episode_Reward/joint_torques_l2: -0.1460 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0161 Episode_Reward/unnecessary_wheel_lift: -0.0886 Episode_Reward/rolling_leg_motion: -0.5815 Episode_Reward/obstacle_progress: +3.8962 Episode_Reward/wheel_body_intrusion: -0.2509 Episode_Reward/pair_body_intrusion: -0.2852 Curriculum/terrain_level_mean: +4.3069 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1340 Episode_Termination/time_out: 0.4937 Episode_Termination/fell_over: 0.0253 Episode_Termination/course_exit: 0.4684 Episode_Termination/illegal_nonwheel_contact: 0.0127 ============================================================================== ============================================================================== Learning iteration 47150/50000 Total steps: 193,126,400 Steps per second: 13,816 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1416.7 Actor loss: -9.5081 Critic loss: +0.0051 World model loss: +0.5628 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.1556 Episode_Reward/track_angular_velocity: +53.4681 Episode_Reward/upright: +27.1300 Episode_Reward/leg_posture: -2.6870 Episode_Reward/standstill: -6.6662 Episode_Reward/action_rate_l2: -0.5952 Episode_Reward/joint_torques_l2: -0.1512 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0166 Episode_Reward/unnecessary_wheel_lift: -0.0790 Episode_Reward/rolling_leg_motion: -0.5857 Episode_Reward/obstacle_progress: +4.0630 Episode_Reward/wheel_body_intrusion: -0.2696 Episode_Reward/pair_body_intrusion: -0.3060 Curriculum/terrain_level_mean: +4.3074 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1356 Episode_Termination/time_out: 0.5390 Episode_Termination/fell_over: 0.0213 Episode_Termination/course_exit: 0.4184 Episode_Termination/illegal_nonwheel_contact: 0.0426 ============================================================================== ============================================================================== Learning iteration 47200/50000 Total steps: 193,331,200 Steps per second: 13,818 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1306.8 Actor loss: -9.5106 Critic loss: +0.0043 World model loss: +0.5546 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2353 Episode_Reward/track_angular_velocity: +49.2766 Episode_Reward/upright: +25.0356 Episode_Reward/leg_posture: -2.5124 Episode_Reward/standstill: -6.1513 Episode_Reward/action_rate_l2: -0.5477 Episode_Reward/joint_torques_l2: -0.1396 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0562 Episode_Reward/unnecessary_wheel_lift: -0.0955 Episode_Reward/rolling_leg_motion: -0.5489 Episode_Reward/obstacle_progress: +3.7595 Episode_Reward/wheel_body_intrusion: -0.2662 Episode_Reward/pair_body_intrusion: -0.2594 Curriculum/terrain_level_mean: +4.3047 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1369 Episode_Termination/time_out: 0.4387 Episode_Termination/fell_over: 0.0645 Episode_Termination/course_exit: 0.4194 Episode_Termination/illegal_nonwheel_contact: 0.0839 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1008] RECOVERY reset 1 invalid environments; first IDs=[2622] ============================================================================== Learning iteration 47250/50000 Total steps: 193,536,000 Steps per second: 13,820 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1385.4 Actor loss: -9.5328 Critic loss: +0.0050 World model loss: +0.5161 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.1819 Episode_Reward/track_angular_velocity: +52.3059 Episode_Reward/upright: +26.5726 Episode_Reward/leg_posture: -2.6394 Episode_Reward/standstill: -6.6116 Episode_Reward/action_rate_l2: -0.5834 Episode_Reward/joint_torques_l2: -0.1468 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0216 Episode_Reward/unnecessary_wheel_lift: -0.1000 Episode_Reward/rolling_leg_motion: -0.5971 Episode_Reward/obstacle_progress: +3.9134 Episode_Reward/wheel_body_intrusion: -0.2654 Episode_Reward/pair_body_intrusion: -0.2873 Curriculum/terrain_level_mean: +4.3057 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4693 Episode_Termination/fell_over: 0.0223 Episode_Termination/course_exit: 0.4860 Episode_Termination/illegal_nonwheel_contact: 0.0391 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2017] ============================================================================== Learning iteration 47300/50000 Total steps: 193,740,800 Steps per second: 13,822 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1351.8 Actor loss: -9.4898 Critic loss: +0.0104 World model loss: +0.5693 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1392 Episode_Reward/track_angular_velocity: +51.1311 Episode_Reward/upright: +25.9260 Episode_Reward/leg_posture: -2.6274 Episode_Reward/standstill: -6.6489 Episode_Reward/action_rate_l2: -0.5598 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0238 Episode_Reward/unnecessary_wheel_lift: -0.0856 Episode_Reward/rolling_leg_motion: -0.5750 Episode_Reward/obstacle_progress: +3.7265 Episode_Reward/wheel_body_intrusion: -0.2650 Episode_Reward/pair_body_intrusion: -0.2667 Curriculum/terrain_level_mean: +4.3049 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0913 Metrics/base_velocity/error_vel_yaw: 0.1332 Episode_Termination/time_out: 0.4564 Episode_Termination/fell_over: 0.0403 Episode_Termination/course_exit: 0.4832 Episode_Termination/illegal_nonwheel_contact: 0.0470 ============================================================================== ============================================================================== Learning iteration 47350/50000 Total steps: 193,945,600 Steps per second: 13,823 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1286.8 Actor loss: -9.5268 Critic loss: +0.0036 World model loss: +0.5638 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.7456 Episode_Reward/track_angular_velocity: +48.6416 Episode_Reward/upright: +24.6939 Episode_Reward/leg_posture: -2.4976 Episode_Reward/standstill: -5.7126 Episode_Reward/action_rate_l2: -0.5275 Episode_Reward/joint_torques_l2: -0.1363 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0126 Episode_Reward/unnecessary_wheel_lift: -0.0631 Episode_Reward/rolling_leg_motion: -0.5434 Episode_Reward/obstacle_progress: +3.5162 Episode_Reward/wheel_body_intrusion: -0.2675 Episode_Reward/pair_body_intrusion: -0.2722 Curriculum/terrain_level_mean: +4.3059 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0920 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.3810 Episode_Termination/fell_over: 0.0317 Episode_Termination/course_exit: 0.5794 Episode_Termination/illegal_nonwheel_contact: 0.0159 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2616] ============================================================================== Learning iteration 47400/50000 Total steps: 194,150,400 Steps per second: 13,825 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1340.6 Actor loss: -9.5306 Critic loss: +0.0049 World model loss: +0.5425 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5260 Episode_Reward/track_angular_velocity: +50.6438 Episode_Reward/upright: +25.6931 Episode_Reward/leg_posture: -2.5901 Episode_Reward/standstill: -5.9995 Episode_Reward/action_rate_l2: -0.5536 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0767 Episode_Reward/rolling_leg_motion: -0.5692 Episode_Reward/obstacle_progress: +3.7061 Episode_Reward/wheel_body_intrusion: -0.2551 Episode_Reward/pair_body_intrusion: -0.3112 Curriculum/terrain_level_mean: +4.3083 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0928 Metrics/base_velocity/error_vel_yaw: 0.1347 Episode_Termination/time_out: 0.4238 Episode_Termination/fell_over: 0.0199 Episode_Termination/course_exit: 0.5232 Episode_Termination/illegal_nonwheel_contact: 0.0331 ============================================================================== ============================================================================== Learning iteration 47450/50000 Total steps: 194,355,200 Steps per second: 13,827 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1336.4 Actor loss: -9.5015 Critic loss: +0.0040 World model loss: +0.5294 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2907 Episode_Reward/track_angular_velocity: +50.4587 Episode_Reward/upright: +25.6284 Episode_Reward/leg_posture: -2.5521 Episode_Reward/standstill: -6.0638 Episode_Reward/action_rate_l2: -0.5661 Episode_Reward/joint_torques_l2: -0.1418 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0804 Episode_Reward/rolling_leg_motion: -0.5666 Episode_Reward/obstacle_progress: +3.7882 Episode_Reward/wheel_body_intrusion: -0.2597 Episode_Reward/pair_body_intrusion: -0.2910 Curriculum/terrain_level_mean: +4.3096 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4733 Episode_Termination/fell_over: 0.0200 Episode_Termination/course_exit: 0.4867 Episode_Termination/illegal_nonwheel_contact: 0.0200 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1162] ============================================================================== Learning iteration 47500/50000 Total steps: 194,560,000 Steps per second: 13,829 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1402.6 Actor loss: -9.5066 Critic loss: +0.0057 World model loss: +0.5516 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7546 Episode_Reward/track_angular_velocity: +52.9100 Episode_Reward/upright: +26.8564 Episode_Reward/leg_posture: -2.6626 Episode_Reward/standstill: -6.3509 Episode_Reward/action_rate_l2: -0.5891 Episode_Reward/joint_torques_l2: -0.1483 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0247 Episode_Reward/unnecessary_wheel_lift: -0.0931 Episode_Reward/rolling_leg_motion: -0.5834 Episode_Reward/obstacle_progress: +3.9887 Episode_Reward/wheel_body_intrusion: -0.2797 Episode_Reward/pair_body_intrusion: -0.2882 Curriculum/terrain_level_mean: +4.3096 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0443 Episode_Termination/course_exit: 0.4430 Episode_Termination/illegal_nonwheel_contact: 0.0190 ============================================================================== ============================================================================== Learning iteration 47550/50000 Total steps: 194,764,800 Steps per second: 13,830 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0872 Mean episode length: 1346.2 Actor loss: -9.5140 Critic loss: +0.0044 World model loss: +0.5484 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.6675 Episode_Reward/track_angular_velocity: +50.7923 Episode_Reward/upright: +25.7989 Episode_Reward/leg_posture: -2.5441 Episode_Reward/standstill: -6.1451 Episode_Reward/action_rate_l2: -0.5632 Episode_Reward/joint_torques_l2: -0.1445 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0237 Episode_Reward/unnecessary_wheel_lift: -0.0932 Episode_Reward/rolling_leg_motion: -0.5474 Episode_Reward/obstacle_progress: +3.7605 Episode_Reward/wheel_body_intrusion: -0.2634 Episode_Reward/pair_body_intrusion: -0.2992 Curriculum/terrain_level_mean: +4.3110 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4615 Episode_Termination/fell_over: 0.0420 Episode_Termination/course_exit: 0.4755 Episode_Termination/illegal_nonwheel_contact: 0.0420 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1473] ============================================================================== Learning iteration 47600/50000 Total steps: 194,969,600 Steps per second: 13,832 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1362.1 Actor loss: -9.5164 Critic loss: +0.0038 World model loss: +0.5483 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0317 Episode_Reward/track_angular_velocity: +51.3868 Episode_Reward/upright: +26.1025 Episode_Reward/leg_posture: -2.6117 Episode_Reward/standstill: -6.3624 Episode_Reward/action_rate_l2: -0.5717 Episode_Reward/joint_torques_l2: -0.1463 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0266 Episode_Reward/unnecessary_wheel_lift: -0.0950 Episode_Reward/rolling_leg_motion: -0.5789 Episode_Reward/obstacle_progress: +3.8251 Episode_Reward/wheel_body_intrusion: -0.2778 Episode_Reward/pair_body_intrusion: -0.3099 Curriculum/terrain_level_mean: +4.3103 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0955 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4936 Episode_Termination/fell_over: 0.0385 Episode_Termination/course_exit: 0.4423 Episode_Termination/illegal_nonwheel_contact: 0.0449 ============================================================================== ============================================================================== Learning iteration 47650/50000 Total steps: 195,174,400 Steps per second: 13,833 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1357.0 Actor loss: -9.5082 Critic loss: +0.0042 World model loss: +0.5322 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0369 Episode_Reward/track_angular_velocity: +51.2035 Episode_Reward/upright: +26.0019 Episode_Reward/leg_posture: -2.6254 Episode_Reward/standstill: -6.1255 Episode_Reward/action_rate_l2: -0.5704 Episode_Reward/joint_torques_l2: -0.1453 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0212 Episode_Reward/unnecessary_wheel_lift: -0.0773 Episode_Reward/rolling_leg_motion: -0.5663 Episode_Reward/obstacle_progress: +3.8653 Episode_Reward/wheel_body_intrusion: -0.2959 Episode_Reward/pair_body_intrusion: -0.3236 Curriculum/terrain_level_mean: +4.3108 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4865 Episode_Termination/fell_over: 0.0270 Episode_Termination/course_exit: 0.4595 Episode_Termination/illegal_nonwheel_contact: 0.0270 ============================================================================== ============================================================================== Learning iteration 47700/50000 Total steps: 195,379,200 Steps per second: 13,835 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1328.6 Actor loss: -9.5163 Critic loss: +0.0057 World model loss: +0.5495 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0666 Episode_Reward/track_angular_velocity: +50.1718 Episode_Reward/upright: +25.4354 Episode_Reward/leg_posture: -2.5402 Episode_Reward/standstill: -6.0423 Episode_Reward/action_rate_l2: -0.5677 Episode_Reward/joint_torques_l2: -0.1402 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0234 Episode_Reward/unnecessary_wheel_lift: -0.0944 Episode_Reward/rolling_leg_motion: -0.5760 Episode_Reward/obstacle_progress: +3.7348 Episode_Reward/wheel_body_intrusion: -0.2568 Episode_Reward/pair_body_intrusion: -0.2528 Curriculum/terrain_level_mean: +4.3113 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1354 Episode_Termination/time_out: 0.4457 Episode_Termination/fell_over: 0.0171 Episode_Termination/course_exit: 0.5029 Episode_Termination/illegal_nonwheel_contact: 0.0400 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1701] ============================================================================== Learning iteration 47750/50000 Total steps: 195,584,000 Steps per second: 13,837 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0874 Mean episode length: 1349.5 Actor loss: -9.5351 Critic loss: +0.0036 World model loss: +0.5401 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8664 Episode_Reward/track_angular_velocity: +50.9193 Episode_Reward/upright: +25.8641 Episode_Reward/leg_posture: -2.5800 Episode_Reward/standstill: -6.1028 Episode_Reward/action_rate_l2: -0.5702 Episode_Reward/joint_torques_l2: -0.1442 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0973 Episode_Reward/rolling_leg_motion: -0.5705 Episode_Reward/obstacle_progress: +3.9313 Episode_Reward/wheel_body_intrusion: -0.2772 Episode_Reward/pair_body_intrusion: -0.2613 Curriculum/terrain_level_mean: +4.3115 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4643 Episode_Termination/fell_over: 0.0429 Episode_Termination/course_exit: 0.4857 Episode_Termination/illegal_nonwheel_contact: 0.0357 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2745] RECOVERY reset 1 invalid environments; first IDs=[1452] ============================================================================== Learning iteration 47800/50000 Total steps: 195,788,800 Steps per second: 13,838 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1314.8 Actor loss: -9.5279 Critic loss: +0.0035 World model loss: +0.5718 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5690 Episode_Reward/track_angular_velocity: +49.5810 Episode_Reward/upright: +25.1962 Episode_Reward/leg_posture: -2.5166 Episode_Reward/standstill: -6.0821 Episode_Reward/action_rate_l2: -0.5601 Episode_Reward/joint_torques_l2: -0.1416 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0231 Episode_Reward/unnecessary_wheel_lift: -0.0830 Episode_Reward/rolling_leg_motion: -0.5709 Episode_Reward/obstacle_progress: +3.8296 Episode_Reward/wheel_body_intrusion: -0.2693 Episode_Reward/pair_body_intrusion: -0.2524 Curriculum/terrain_level_mean: +4.3125 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4354 Episode_Termination/fell_over: 0.0340 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0408 ============================================================================== ============================================================================== Learning iteration 47850/50000 Total steps: 195,993,600 Steps per second: 13,840 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1365.2 Actor loss: -9.5618 Critic loss: +0.0037 World model loss: +0.5626 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.4434 Episode_Reward/track_angular_velocity: +51.5668 Episode_Reward/upright: +26.1772 Episode_Reward/leg_posture: -2.6104 Episode_Reward/standstill: -6.5073 Episode_Reward/action_rate_l2: -0.5695 Episode_Reward/joint_torques_l2: -0.1456 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0347 Episode_Reward/unnecessary_wheel_lift: -0.0753 Episode_Reward/rolling_leg_motion: -0.5638 Episode_Reward/obstacle_progress: +3.9224 Episode_Reward/wheel_body_intrusion: -0.2816 Episode_Reward/pair_body_intrusion: -0.2728 Curriculum/terrain_level_mean: +4.3108 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1349 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0476 Episode_Termination/course_exit: 0.4082 Episode_Termination/illegal_nonwheel_contact: 0.0476 ============================================================================== ============================================================================== Learning iteration 47900/50000 Total steps: 196,198,400 Steps per second: 13,842 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0871 Mean episode length: 1378.8 Actor loss: -9.5131 Critic loss: +0.0063 World model loss: +0.5855 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0342 Episode_Reward/track_angular_velocity: +52.1022 Episode_Reward/upright: +26.4441 Episode_Reward/leg_posture: -2.6709 Episode_Reward/standstill: -6.7417 Episode_Reward/action_rate_l2: -0.5813 Episode_Reward/joint_torques_l2: -0.1469 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.0776 Episode_Reward/rolling_leg_motion: -0.5798 Episode_Reward/obstacle_progress: +3.9642 Episode_Reward/wheel_body_intrusion: -0.2800 Episode_Reward/pair_body_intrusion: -0.2912 Curriculum/terrain_level_mean: +4.3123 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4899 Episode_Termination/fell_over: 0.0067 Episode_Termination/course_exit: 0.4765 Episode_Termination/illegal_nonwheel_contact: 0.0336 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[760] ============================================================================== Learning iteration 47950/50000 Total steps: 196,403,200 Steps per second: 13,844 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1356.1 Actor loss: -9.5163 Critic loss: +0.0086 World model loss: +0.5581 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1769 Episode_Reward/track_angular_velocity: +51.2085 Episode_Reward/upright: +26.0408 Episode_Reward/leg_posture: -2.5716 Episode_Reward/standstill: -6.6703 Episode_Reward/action_rate_l2: -0.5846 Episode_Reward/joint_torques_l2: -0.1434 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.1124 Episode_Reward/rolling_leg_motion: -0.5940 Episode_Reward/obstacle_progress: +3.8709 Episode_Reward/wheel_body_intrusion: -0.2678 Episode_Reward/pair_body_intrusion: -0.3034 Curriculum/terrain_level_mean: +4.3132 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0925 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.5111 Episode_Termination/fell_over: 0.0167 Episode_Termination/course_exit: 0.4500 Episode_Termination/illegal_nonwheel_contact: 0.0278 ============================================================================== ============================================================================== Learning iteration 48000/50000 Total steps: 196,608,000 Steps per second: 13,845 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1361.4 Actor loss: -9.5301 Critic loss: +0.0035 World model loss: +0.5916 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5578 Episode_Reward/track_angular_velocity: +51.4571 Episode_Reward/upright: +26.1866 Episode_Reward/leg_posture: -2.5756 Episode_Reward/standstill: -6.4097 Episode_Reward/action_rate_l2: -0.5723 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0213 Episode_Reward/unnecessary_wheel_lift: -0.1046 Episode_Reward/rolling_leg_motion: -0.5862 Episode_Reward/obstacle_progress: +3.8754 Episode_Reward/wheel_body_intrusion: -0.2731 Episode_Reward/pair_body_intrusion: -0.2797 Curriculum/terrain_level_mean: +4.3130 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0906 Metrics/base_velocity/error_vel_yaw: 0.1342 Episode_Termination/time_out: 0.5137 Episode_Termination/fell_over: 0.0342 Episode_Termination/course_exit: 0.4315 Episode_Termination/illegal_nonwheel_contact: 0.0411 ============================================================================== ============================================================================== Learning iteration 48050/50000 Total steps: 196,812,800 Steps per second: 13,847 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0874 Mean episode length: 1351.8 Actor loss: -9.5511 Critic loss: +0.0031 World model loss: +0.5748 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0865 Episode_Reward/track_angular_velocity: +51.1079 Episode_Reward/upright: +26.0009 Episode_Reward/leg_posture: -2.5713 Episode_Reward/standstill: -6.4272 Episode_Reward/action_rate_l2: -0.5618 Episode_Reward/joint_torques_l2: -0.1427 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0132 Episode_Reward/unnecessary_wheel_lift: -0.0790 Episode_Reward/rolling_leg_motion: -0.5563 Episode_Reward/obstacle_progress: +3.8223 Episode_Reward/wheel_body_intrusion: -0.2739 Episode_Reward/pair_body_intrusion: -0.2766 Curriculum/terrain_level_mean: +4.3152 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0916 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4604 Episode_Termination/fell_over: 0.0288 Episode_Termination/course_exit: 0.5036 Episode_Termination/illegal_nonwheel_contact: 0.0144 ============================================================================== ============================================================================== Learning iteration 48100/50000 Total steps: 197,017,600 Steps per second: 13,849 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0869 Mean episode length: 1373.1 Actor loss: -9.5093 Critic loss: +0.0048 World model loss: +0.5273 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8377 Episode_Reward/track_angular_velocity: +51.8714 Episode_Reward/upright: +26.3593 Episode_Reward/leg_posture: -2.6738 Episode_Reward/standstill: -6.8659 Episode_Reward/action_rate_l2: -0.5647 Episode_Reward/joint_torques_l2: -0.1461 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0896 Episode_Reward/rolling_leg_motion: -0.5759 Episode_Reward/obstacle_progress: +3.8686 Episode_Reward/wheel_body_intrusion: -0.2924 Episode_Reward/pair_body_intrusion: -0.3255 Curriculum/terrain_level_mean: +4.3145 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4937 Episode_Termination/fell_over: 0.0380 Episode_Termination/course_exit: 0.4620 Episode_Termination/illegal_nonwheel_contact: 0.0190 ============================================================================== ============================================================================== Learning iteration 48150/50000 Total steps: 197,222,400 Steps per second: 13,851 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0869 Mean episode length: 1346.7 Actor loss: -9.5065 Critic loss: +0.0050 World model loss: +0.5570 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.9058 Episode_Reward/track_angular_velocity: +50.9136 Episode_Reward/upright: +25.8672 Episode_Reward/leg_posture: -2.6386 Episode_Reward/standstill: -6.7728 Episode_Reward/action_rate_l2: -0.5515 Episode_Reward/joint_torques_l2: -0.1423 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0210 Episode_Reward/unnecessary_wheel_lift: -0.0910 Episode_Reward/rolling_leg_motion: -0.5770 Episode_Reward/obstacle_progress: +3.7406 Episode_Reward/wheel_body_intrusion: -0.2726 Episode_Reward/pair_body_intrusion: -0.3133 Curriculum/terrain_level_mean: +4.3152 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0919 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.4762 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4694 Episode_Termination/illegal_nonwheel_contact: 0.0408 ============================================================================== ============================================================================== Learning iteration 48200/50000 Total steps: 197,427,200 Steps per second: 13,852 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.183s Mean reward: +0.0874 Mean episode length: 1319.2 Actor loss: -9.5143 Critic loss: +0.0044 World model loss: +0.5207 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.1282 Episode_Reward/track_angular_velocity: +49.8837 Episode_Reward/upright: +25.3655 Episode_Reward/leg_posture: -2.5524 Episode_Reward/standstill: -6.2833 Episode_Reward/action_rate_l2: -0.5421 Episode_Reward/joint_torques_l2: -0.1386 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0283 Episode_Reward/unnecessary_wheel_lift: -0.0821 Episode_Reward/rolling_leg_motion: -0.5719 Episode_Reward/obstacle_progress: +3.7218 Episode_Reward/wheel_body_intrusion: -0.2695 Episode_Reward/pair_body_intrusion: -0.3004 Curriculum/terrain_level_mean: +4.3164 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0899 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4277 Episode_Termination/fell_over: 0.0440 Episode_Termination/course_exit: 0.5031 Episode_Termination/illegal_nonwheel_contact: 0.0377 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2097] ============================================================================== Learning iteration 48250/50000 Total steps: 197,632,000 Steps per second: 13,854 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0876 Mean episode length: 1393.5 Actor loss: -9.5666 Critic loss: +0.0058 World model loss: +0.5157 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.7315 Episode_Reward/track_angular_velocity: +52.6849 Episode_Reward/upright: +26.7853 Episode_Reward/leg_posture: -2.6836 Episode_Reward/standstill: -6.4157 Episode_Reward/action_rate_l2: -0.5747 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0163 Episode_Reward/unnecessary_wheel_lift: -0.0828 Episode_Reward/rolling_leg_motion: -0.5910 Episode_Reward/obstacle_progress: +3.9447 Episode_Reward/wheel_body_intrusion: -0.2953 Episode_Reward/pair_body_intrusion: -0.3185 Curriculum/terrain_level_mean: +4.3179 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0905 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.4710 Episode_Termination/fell_over: 0.0129 Episode_Termination/course_exit: 0.4903 Episode_Termination/illegal_nonwheel_contact: 0.0258 ============================================================================== ============================================================================== Learning iteration 48300/50000 Total steps: 197,836,800 Steps per second: 13,856 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1349.2 Actor loss: -9.5140 Critic loss: +0.0040 World model loss: +0.5557 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0061 Episode_Reward/track_angular_velocity: +50.9293 Episode_Reward/upright: +25.9143 Episode_Reward/leg_posture: -2.5735 Episode_Reward/standstill: -5.9757 Episode_Reward/action_rate_l2: -0.5771 Episode_Reward/joint_torques_l2: -0.1432 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0865 Episode_Reward/rolling_leg_motion: -0.5879 Episode_Reward/obstacle_progress: +3.7722 Episode_Reward/wheel_body_intrusion: -0.2704 Episode_Reward/pair_body_intrusion: -0.2807 Curriculum/terrain_level_mean: +4.3198 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0915 Metrics/base_velocity/error_vel_yaw: 0.1357 Episode_Termination/time_out: 0.4670 Episode_Termination/fell_over: 0.0165 Episode_Termination/course_exit: 0.4945 Episode_Termination/illegal_nonwheel_contact: 0.0275 ============================================================================== ============================================================================== Learning iteration 48350/50000 Total steps: 198,041,600 Steps per second: 13,857 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0876 Mean episode length: 1270.9 Actor loss: -9.5192 Critic loss: +0.0040 World model loss: +0.5315 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +45.2727 Episode_Reward/track_angular_velocity: +47.9953 Episode_Reward/upright: +24.4124 Episode_Reward/leg_posture: -2.4302 Episode_Reward/standstill: -5.7118 Episode_Reward/action_rate_l2: -0.5259 Episode_Reward/joint_torques_l2: -0.1342 Episode_Reward/dof_pos_limits: -0.0014 Episode_Reward/terminated: -0.0248 Episode_Reward/unnecessary_wheel_lift: -0.0780 Episode_Reward/rolling_leg_motion: -0.5491 Episode_Reward/obstacle_progress: +3.5452 Episode_Reward/wheel_body_intrusion: -0.2640 Episode_Reward/pair_body_intrusion: -0.2425 Curriculum/terrain_level_mean: +4.3193 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0912 Metrics/base_velocity/error_vel_yaw: 0.1351 Episode_Termination/time_out: 0.3987 Episode_Termination/fell_over: 0.0316 Episode_Termination/course_exit: 0.5380 Episode_Termination/illegal_nonwheel_contact: 0.0443 ============================================================================== ============================================================================== Learning iteration 48400/50000 Total steps: 198,246,400 Steps per second: 13,859 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1330.2 Actor loss: -9.5401 Critic loss: +0.0038 World model loss: +0.5538 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2472 Episode_Reward/track_angular_velocity: +50.2063 Episode_Reward/upright: +25.5439 Episode_Reward/leg_posture: -2.5434 Episode_Reward/standstill: -5.9589 Episode_Reward/action_rate_l2: -0.5592 Episode_Reward/joint_torques_l2: -0.1402 Episode_Reward/dof_pos_limits: -0.0016 Episode_Reward/terminated: -0.0078 Episode_Reward/unnecessary_wheel_lift: -0.0797 Episode_Reward/rolling_leg_motion: -0.5690 Episode_Reward/obstacle_progress: +3.8053 Episode_Reward/wheel_body_intrusion: -0.2607 Episode_Reward/pair_body_intrusion: -0.2643 Curriculum/terrain_level_mean: +4.3223 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0926 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.4155 Episode_Termination/fell_over: 0.0141 Episode_Termination/course_exit: 0.5634 Episode_Termination/illegal_nonwheel_contact: 0.0070 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2204] ============================================================================== Learning iteration 48450/50000 Total steps: 198,451,200 Steps per second: 13,861 Iteration time: 0.296s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0871 Mean episode length: 1345.1 Actor loss: -9.5488 Critic loss: +0.0036 World model loss: +0.5357 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.8634 Episode_Reward/track_angular_velocity: +50.7560 Episode_Reward/upright: +25.8212 Episode_Reward/leg_posture: -2.5184 Episode_Reward/standstill: -6.6832 Episode_Reward/action_rate_l2: -0.5658 Episode_Reward/joint_torques_l2: -0.1415 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0368 Episode_Reward/unnecessary_wheel_lift: -0.1219 Episode_Reward/rolling_leg_motion: -0.5716 Episode_Reward/obstacle_progress: +3.8897 Episode_Reward/wheel_body_intrusion: -0.2526 Episode_Reward/pair_body_intrusion: -0.2937 Curriculum/terrain_level_mean: +4.3196 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0918 Metrics/base_velocity/error_vel_yaw: 0.1363 Episode_Termination/time_out: 0.4737 Episode_Termination/fell_over: 0.0602 Episode_Termination/course_exit: 0.4211 Episode_Termination/illegal_nonwheel_contact: 0.0677 ============================================================================== ============================================================================== Learning iteration 48500/50000 Total steps: 198,656,000 Steps per second: 13,862 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1406.2 Actor loss: -9.5250 Critic loss: +0.0044 World model loss: +0.5416 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.9642 Episode_Reward/track_angular_velocity: +53.0737 Episode_Reward/upright: +26.9631 Episode_Reward/leg_posture: -2.6749 Episode_Reward/standstill: -6.6358 Episode_Reward/action_rate_l2: -0.5739 Episode_Reward/joint_torques_l2: -0.1489 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0165 Episode_Reward/unnecessary_wheel_lift: -0.1091 Episode_Reward/rolling_leg_motion: -0.5823 Episode_Reward/obstacle_progress: +4.0720 Episode_Reward/wheel_body_intrusion: -0.2853 Episode_Reward/pair_body_intrusion: -0.2927 Curriculum/terrain_level_mean: +4.3198 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1358 Episode_Termination/time_out: 0.5033 Episode_Termination/fell_over: 0.0199 Episode_Termination/course_exit: 0.4570 Episode_Termination/illegal_nonwheel_contact: 0.0331 ============================================================================== ============================================================================== Learning iteration 48550/50000 Total steps: 198,860,800 Steps per second: 13,864 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1300.0 Actor loss: -9.5453 Critic loss: +0.0031 World model loss: +0.5477 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.2313 Episode_Reward/track_angular_velocity: +49.1258 Episode_Reward/upright: +24.9471 Episode_Reward/leg_posture: -2.4713 Episode_Reward/standstill: -5.7410 Episode_Reward/action_rate_l2: -0.5433 Episode_Reward/joint_torques_l2: -0.1375 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0305 Episode_Reward/unnecessary_wheel_lift: -0.0885 Episode_Reward/rolling_leg_motion: -0.5559 Episode_Reward/obstacle_progress: +3.7746 Episode_Reward/wheel_body_intrusion: -0.2583 Episode_Reward/pair_body_intrusion: -0.2793 Curriculum/terrain_level_mean: +4.3208 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0921 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.4412 Episode_Termination/fell_over: 0.0294 Episode_Termination/course_exit: 0.4882 Episode_Termination/illegal_nonwheel_contact: 0.0471 ============================================================================== ============================================================================== Learning iteration 48600/50000 Total steps: 199,065,600 Steps per second: 13,866 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0877 Mean episode length: 1347.7 Actor loss: -9.5336 Critic loss: +0.0032 World model loss: +0.5348 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0040 Episode_Reward/track_angular_velocity: +50.9309 Episode_Reward/upright: +25.8526 Episode_Reward/leg_posture: -2.5473 Episode_Reward/standstill: -5.9906 Episode_Reward/action_rate_l2: -0.5651 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0184 Episode_Reward/unnecessary_wheel_lift: -0.0895 Episode_Reward/rolling_leg_motion: -0.5688 Episode_Reward/obstacle_progress: +3.9121 Episode_Reward/wheel_body_intrusion: -0.2564 Episode_Reward/pair_body_intrusion: -0.2776 Curriculum/terrain_level_mean: +4.3223 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0914 Metrics/base_velocity/error_vel_yaw: 0.1346 Episode_Termination/time_out: 0.4564 Episode_Termination/fell_over: 0.0201 Episode_Termination/course_exit: 0.5034 Episode_Termination/illegal_nonwheel_contact: 0.0336 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_048600.mp4 RECOVERY reset 1 invalid environments; first IDs=[936] ============================================================================== Learning iteration 48650/50000 Total steps: 199,270,400 Steps per second: 13,862 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0874 Mean episode length: 1333.1 Actor loss: -9.5099 Critic loss: +0.0038 World model loss: +0.5452 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2447 Episode_Reward/track_angular_velocity: +50.3630 Episode_Reward/upright: +25.5577 Episode_Reward/leg_posture: -2.4958 Episode_Reward/standstill: -6.1407 Episode_Reward/action_rate_l2: -0.5663 Episode_Reward/joint_torques_l2: -0.1406 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0260 Episode_Reward/unnecessary_wheel_lift: -0.0852 Episode_Reward/rolling_leg_motion: -0.5709 Episode_Reward/obstacle_progress: +3.8588 Episode_Reward/wheel_body_intrusion: -0.2628 Episode_Reward/pair_body_intrusion: -0.2732 Curriculum/terrain_level_mean: +4.3237 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4014 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.5306 Episode_Termination/illegal_nonwheel_contact: 0.0544 ============================================================================== ============================================================================== Learning iteration 48700/50000 Total steps: 199,475,200 Steps per second: 13,863 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0872 Mean episode length: 1372.1 Actor loss: -9.5293 Critic loss: +0.0124 World model loss: +0.5491 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.6796 Episode_Reward/track_angular_velocity: +51.8179 Episode_Reward/upright: +26.2895 Episode_Reward/leg_posture: -2.5686 Episode_Reward/standstill: -6.6186 Episode_Reward/action_rate_l2: -0.5823 Episode_Reward/joint_torques_l2: -0.1450 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0214 Episode_Reward/unnecessary_wheel_lift: -0.0813 Episode_Reward/rolling_leg_motion: -0.5982 Episode_Reward/obstacle_progress: +4.0388 Episode_Reward/wheel_body_intrusion: -0.2622 Episode_Reward/pair_body_intrusion: -0.2858 Curriculum/terrain_level_mean: +4.3245 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4599 Episode_Termination/fell_over: 0.0292 Episode_Termination/course_exit: 0.4818 Episode_Termination/illegal_nonwheel_contact: 0.0438 ============================================================================== ============================================================================== Learning iteration 48750/50000 Total steps: 199,680,000 Steps per second: 13,865 Iteration time: 0.295s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1358.8 Actor loss: -9.5445 Critic loss: +0.0036 World model loss: +0.5544 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.1063 Episode_Reward/track_angular_velocity: +51.2686 Episode_Reward/upright: +26.0420 Episode_Reward/leg_posture: -2.5208 Episode_Reward/standstill: -6.2660 Episode_Reward/action_rate_l2: -0.5722 Episode_Reward/joint_torques_l2: -0.1444 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0265 Episode_Reward/unnecessary_wheel_lift: -0.0723 Episode_Reward/rolling_leg_motion: -0.5559 Episode_Reward/obstacle_progress: +3.9636 Episode_Reward/wheel_body_intrusion: -0.2602 Episode_Reward/pair_body_intrusion: -0.2948 Curriculum/terrain_level_mean: +4.3235 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.5000 Episode_Termination/fell_over: 0.0164 Episode_Termination/course_exit: 0.4180 Episode_Termination/illegal_nonwheel_contact: 0.0656 ============================================================================== ============================================================================== Learning iteration 48800/50000 Total steps: 199,884,800 Steps per second: 13,867 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0870 Mean episode length: 1348.1 Actor loss: -9.5034 Critic loss: +0.0077 World model loss: +0.5298 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.7753 Episode_Reward/track_angular_velocity: +50.8805 Episode_Reward/upright: +25.8086 Episode_Reward/leg_posture: -2.5442 Episode_Reward/standstill: -6.5966 Episode_Reward/action_rate_l2: -0.5681 Episode_Reward/joint_torques_l2: -0.1437 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0290 Episode_Reward/unnecessary_wheel_lift: -0.1144 Episode_Reward/rolling_leg_motion: -0.5617 Episode_Reward/obstacle_progress: +3.9282 Episode_Reward/wheel_body_intrusion: -0.2684 Episode_Reward/pair_body_intrusion: -0.2684 Curriculum/terrain_level_mean: +4.3247 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0936 Metrics/base_velocity/error_vel_yaw: 0.1362 Episode_Termination/time_out: 0.4817 Episode_Termination/fell_over: 0.0427 Episode_Termination/course_exit: 0.4512 Episode_Termination/illegal_nonwheel_contact: 0.0427 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1750] ============================================================================== Learning iteration 48850/50000 Total steps: 200,089,600 Steps per second: 13,868 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0869 Mean episode length: 1340.8 Actor loss: -9.5229 Critic loss: +0.0077 World model loss: +0.5502 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.4639 Episode_Reward/track_angular_velocity: +50.6311 Episode_Reward/upright: +25.6906 Episode_Reward/leg_posture: -2.5351 Episode_Reward/standstill: -6.5441 Episode_Reward/action_rate_l2: -0.5529 Episode_Reward/joint_torques_l2: -0.1430 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0860 Episode_Reward/rolling_leg_motion: -0.5747 Episode_Reward/obstacle_progress: +3.8030 Episode_Reward/wheel_body_intrusion: -0.2793 Episode_Reward/pair_body_intrusion: -0.2957 Curriculum/terrain_level_mean: +4.3267 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4803 Episode_Termination/fell_over: 0.0197 Episode_Termination/course_exit: 0.4803 Episode_Termination/illegal_nonwheel_contact: 0.0329 ============================================================================== ============================================================================== Learning iteration 48900/50000 Total steps: 200,294,400 Steps per second: 13,870 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1329.6 Actor loss: -9.5464 Critic loss: +0.0040 World model loss: +0.5160 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2345 Episode_Reward/track_angular_velocity: +50.2661 Episode_Reward/upright: +25.4965 Episode_Reward/leg_posture: -2.5168 Episode_Reward/standstill: -6.0054 Episode_Reward/action_rate_l2: -0.5523 Episode_Reward/joint_torques_l2: -0.1429 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0312 Episode_Reward/unnecessary_wheel_lift: -0.0741 Episode_Reward/rolling_leg_motion: -0.5613 Episode_Reward/obstacle_progress: +3.8612 Episode_Reward/wheel_body_intrusion: -0.2836 Episode_Reward/pair_body_intrusion: -0.2725 Curriculum/terrain_level_mean: +4.3254 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1335 Episode_Termination/time_out: 0.4476 Episode_Termination/fell_over: 0.0559 Episode_Termination/course_exit: 0.4685 Episode_Termination/illegal_nonwheel_contact: 0.0559 ============================================================================== ============================================================================== Learning iteration 48950/50000 Total steps: 200,499,200 Steps per second: 13,872 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1421.5 Actor loss: -9.5699 Critic loss: +0.0043 World model loss: +0.5342 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.4493 Episode_Reward/track_angular_velocity: +53.7212 Episode_Reward/upright: +27.2718 Episode_Reward/leg_posture: -2.6457 Episode_Reward/standstill: -6.4853 Episode_Reward/action_rate_l2: -0.5910 Episode_Reward/joint_torques_l2: -0.1509 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0250 Episode_Reward/unnecessary_wheel_lift: -0.0886 Episode_Reward/rolling_leg_motion: -0.6065 Episode_Reward/obstacle_progress: +4.0446 Episode_Reward/wheel_body_intrusion: -0.2890 Episode_Reward/pair_body_intrusion: -0.2807 Curriculum/terrain_level_mean: +4.3254 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0932 Metrics/base_velocity/error_vel_yaw: 0.1343 Episode_Termination/time_out: 0.5190 Episode_Termination/fell_over: 0.0253 Episode_Termination/course_exit: 0.4241 Episode_Termination/illegal_nonwheel_contact: 0.0380 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2389] ============================================================================== Learning iteration 49000/50000 Total steps: 200,704,000 Steps per second: 13,873 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.192s Mean reward: +0.0877 Mean episode length: 1351.5 Actor loss: -9.5338 Critic loss: +0.0040 World model loss: +0.5691 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.0078 Episode_Reward/track_angular_velocity: +51.1073 Episode_Reward/upright: +25.9315 Episode_Reward/leg_posture: -2.4811 Episode_Reward/standstill: -5.8944 Episode_Reward/action_rate_l2: -0.5546 Episode_Reward/joint_torques_l2: -0.1428 Episode_Reward/dof_pos_limits: -0.0010 Episode_Reward/terminated: -0.0185 Episode_Reward/unnecessary_wheel_lift: -0.0853 Episode_Reward/rolling_leg_motion: -0.5978 Episode_Reward/obstacle_progress: +3.7789 Episode_Reward/wheel_body_intrusion: -0.2605 Episode_Reward/pair_body_intrusion: -0.2891 Curriculum/terrain_level_mean: +4.3267 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1339 Episode_Termination/time_out: 0.5060 Episode_Termination/fell_over: 0.0417 Episode_Termination/course_exit: 0.4524 Episode_Termination/illegal_nonwheel_contact: 0.0060 ============================================================================== ============================================================================== Learning iteration 49050/50000 Total steps: 200,908,800 Steps per second: 13,875 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1329.9 Actor loss: -9.5229 Critic loss: +0.0054 World model loss: +0.5478 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.2644 Episode_Reward/track_angular_velocity: +50.2447 Episode_Reward/upright: +25.4793 Episode_Reward/leg_posture: -2.4812 Episode_Reward/standstill: -5.6937 Episode_Reward/action_rate_l2: -0.5476 Episode_Reward/joint_torques_l2: -0.1410 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0234 Episode_Reward/unnecessary_wheel_lift: -0.0964 Episode_Reward/rolling_leg_motion: -0.5769 Episode_Reward/obstacle_progress: +3.7343 Episode_Reward/wheel_body_intrusion: -0.2750 Episode_Reward/pair_body_intrusion: -0.2896 Curriculum/terrain_level_mean: +4.3281 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0922 Metrics/base_velocity/error_vel_yaw: 0.1344 Episode_Termination/time_out: 0.4336 Episode_Termination/fell_over: 0.0210 Episode_Termination/course_exit: 0.5035 Episode_Termination/illegal_nonwheel_contact: 0.0420 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1326] ============================================================================== Learning iteration 49100/50000 Total steps: 201,113,600 Steps per second: 13,876 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1259.0 Actor loss: -9.5204 Critic loss: +0.0058 World model loss: +0.5785 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +44.6196 Episode_Reward/track_angular_velocity: +47.5697 Episode_Reward/upright: +24.1461 Episode_Reward/leg_posture: -2.3457 Episode_Reward/standstill: -5.4179 Episode_Reward/action_rate_l2: -0.5228 Episode_Reward/joint_torques_l2: -0.1335 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0246 Episode_Reward/unnecessary_wheel_lift: -0.0703 Episode_Reward/rolling_leg_motion: -0.5292 Episode_Reward/obstacle_progress: +3.5579 Episode_Reward/wheel_body_intrusion: -0.2333 Episode_Reward/pair_body_intrusion: -0.2618 Curriculum/terrain_level_mean: +4.3284 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1345 Episode_Termination/time_out: 0.3497 Episode_Termination/fell_over: 0.0350 Episode_Termination/course_exit: 0.5804 Episode_Termination/illegal_nonwheel_contact: 0.0490 ============================================================================== ============================================================================== Learning iteration 49150/50000 Total steps: 201,318,400 Steps per second: 13,878 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0876 Mean episode length: 1310.1 Actor loss: -9.5244 Critic loss: +0.0064 World model loss: +0.5099 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.3792 Episode_Reward/track_angular_velocity: +49.4754 Episode_Reward/upright: +25.1065 Episode_Reward/leg_posture: -2.4249 Episode_Reward/standstill: -5.7092 Episode_Reward/action_rate_l2: -0.5520 Episode_Reward/joint_torques_l2: -0.1390 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0230 Episode_Reward/unnecessary_wheel_lift: -0.0809 Episode_Reward/rolling_leg_motion: -0.5518 Episode_Reward/obstacle_progress: +3.7313 Episode_Reward/wheel_body_intrusion: -0.2382 Episode_Reward/pair_body_intrusion: -0.2643 Curriculum/terrain_level_mean: +4.3286 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1352 Episode_Termination/time_out: 0.4326 Episode_Termination/fell_over: 0.0142 Episode_Termination/course_exit: 0.5035 Episode_Termination/illegal_nonwheel_contact: 0.0567 ============================================================================== ============================================================================== Learning iteration 49200/50000 Total steps: 201,523,200 Steps per second: 13,880 Iteration time: 0.295s Collection time: 0.075s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1382.2 Actor loss: -9.5145 Critic loss: +0.0111 World model loss: +0.5686 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9344 Episode_Reward/track_angular_velocity: +52.1596 Episode_Reward/upright: +26.4660 Episode_Reward/leg_posture: -2.5641 Episode_Reward/standstill: -6.3275 Episode_Reward/action_rate_l2: -0.5843 Episode_Reward/joint_torques_l2: -0.1481 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0135 Episode_Reward/unnecessary_wheel_lift: -0.1006 Episode_Reward/rolling_leg_motion: -0.5787 Episode_Reward/obstacle_progress: +3.9507 Episode_Reward/wheel_body_intrusion: -0.2820 Episode_Reward/pair_body_intrusion: -0.3226 Curriculum/terrain_level_mean: +4.3301 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0942 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4295 Episode_Termination/fell_over: 0.0067 Episode_Termination/course_exit: 0.5369 Episode_Termination/illegal_nonwheel_contact: 0.0268 ============================================================================== ============================================================================== Learning iteration 49250/50000 Total steps: 201,728,000 Steps per second: 13,881 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1316.3 Actor loss: -9.4948 Critic loss: +0.0046 World model loss: +0.5345 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5452 Episode_Reward/track_angular_velocity: +49.6297 Episode_Reward/upright: +25.1910 Episode_Reward/leg_posture: -2.4896 Episode_Reward/standstill: -5.8877 Episode_Reward/action_rate_l2: -0.5514 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0013 Episode_Reward/terminated: -0.0386 Episode_Reward/unnecessary_wheel_lift: -0.0965 Episode_Reward/rolling_leg_motion: -0.5620 Episode_Reward/obstacle_progress: +3.7446 Episode_Reward/wheel_body_intrusion: -0.2623 Episode_Reward/pair_body_intrusion: -0.2568 Curriculum/terrain_level_mean: +4.3296 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0945 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4444 Episode_Termination/fell_over: 0.0117 Episode_Termination/course_exit: 0.4678 Episode_Termination/illegal_nonwheel_contact: 0.0819 ============================================================================== ============================================================================== Learning iteration 49300/50000 Total steps: 201,932,800 Steps per second: 13,883 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1337.6 Actor loss: -9.5161 Critic loss: +0.0053 World model loss: +0.5746 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3823 Episode_Reward/track_angular_velocity: +50.4490 Episode_Reward/upright: +25.6411 Episode_Reward/leg_posture: -2.5146 Episode_Reward/standstill: -6.0161 Episode_Reward/action_rate_l2: -0.5596 Episode_Reward/joint_torques_l2: -0.1420 Episode_Reward/dof_pos_limits: -0.0004 Episode_Reward/terminated: -0.0209 Episode_Reward/unnecessary_wheel_lift: -0.0969 Episode_Reward/rolling_leg_motion: -0.5741 Episode_Reward/obstacle_progress: +3.7802 Episode_Reward/wheel_body_intrusion: -0.2538 Episode_Reward/pair_body_intrusion: -0.2927 Curriculum/terrain_level_mean: +4.3308 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0939 Metrics/base_velocity/error_vel_yaw: 0.1367 Episode_Termination/time_out: 0.4625 Episode_Termination/fell_over: 0.0312 Episode_Termination/course_exit: 0.4875 Episode_Termination/illegal_nonwheel_contact: 0.0250 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1098] ============================================================================== Learning iteration 49350/50000 Total steps: 202,137,600 Steps per second: 13,885 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0873 Mean episode length: 1342.3 Actor loss: -9.5245 Critic loss: +0.0087 World model loss: +0.5296 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.5814 Episode_Reward/track_angular_velocity: +50.5739 Episode_Reward/upright: +25.7450 Episode_Reward/leg_posture: -2.5264 Episode_Reward/standstill: -6.0654 Episode_Reward/action_rate_l2: -0.5660 Episode_Reward/joint_torques_l2: -0.1435 Episode_Reward/dof_pos_limits: -0.0002 Episode_Reward/terminated: -0.0157 Episode_Reward/unnecessary_wheel_lift: -0.0772 Episode_Reward/rolling_leg_motion: -0.5757 Episode_Reward/obstacle_progress: +3.8152 Episode_Reward/wheel_body_intrusion: -0.2631 Episode_Reward/pair_body_intrusion: -0.2711 Curriculum/terrain_level_mean: +4.3308 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1381 Episode_Termination/time_out: 0.4823 Episode_Termination/fell_over: 0.0142 Episode_Termination/course_exit: 0.4752 Episode_Termination/illegal_nonwheel_contact: 0.0284 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[778] ============================================================================== Learning iteration 49400/50000 Total steps: 202,342,400 Steps per second: 13,886 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0872 Mean episode length: 1382.1 Actor loss: -9.5452 Critic loss: +0.0057 World model loss: +0.5156 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.9784 Episode_Reward/track_angular_velocity: +52.0518 Episode_Reward/upright: +26.4826 Episode_Reward/leg_posture: -2.6122 Episode_Reward/standstill: -6.3842 Episode_Reward/action_rate_l2: -0.5893 Episode_Reward/joint_torques_l2: -0.1475 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0162 Episode_Reward/unnecessary_wheel_lift: -0.0941 Episode_Reward/rolling_leg_motion: -0.5844 Episode_Reward/obstacle_progress: +3.9430 Episode_Reward/wheel_body_intrusion: -0.2663 Episode_Reward/pair_body_intrusion: -0.2996 Curriculum/terrain_level_mean: +4.3323 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1387 Episode_Termination/time_out: 0.4843 Episode_Termination/fell_over: 0.0189 Episode_Termination/course_exit: 0.4780 Episode_Termination/illegal_nonwheel_contact: 0.0314 ============================================================================== ============================================================================== Learning iteration 49450/50000 Total steps: 202,547,200 Steps per second: 13,888 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0873 Mean episode length: 1395.2 Actor loss: -9.5119 Critic loss: +0.0052 World model loss: +0.5660 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.5187 Episode_Reward/track_angular_velocity: +52.6016 Episode_Reward/upright: +26.7404 Episode_Reward/leg_posture: -2.6243 Episode_Reward/standstill: -6.4191 Episode_Reward/action_rate_l2: -0.5877 Episode_Reward/joint_torques_l2: -0.1490 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0436 Episode_Reward/unnecessary_wheel_lift: -0.1126 Episode_Reward/rolling_leg_motion: -0.5980 Episode_Reward/obstacle_progress: +3.9916 Episode_Reward/wheel_body_intrusion: -0.2582 Episode_Reward/pair_body_intrusion: -0.2864 Curriculum/terrain_level_mean: +4.3306 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0931 Metrics/base_velocity/error_vel_yaw: 0.1374 Episode_Termination/time_out: 0.4915 Episode_Termination/fell_over: 0.0395 Episode_Termination/course_exit: 0.4181 Episode_Termination/illegal_nonwheel_contact: 0.0791 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2374] ============================================================================== Learning iteration 49500/50000 Total steps: 202,752,000 Steps per second: 13,889 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.184s Mean reward: +0.0875 Mean episode length: 1313.4 Actor loss: -9.5232 Critic loss: +0.0064 World model loss: +0.5403 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.6585 Episode_Reward/track_angular_velocity: +49.5539 Episode_Reward/upright: +25.1485 Episode_Reward/leg_posture: -2.5033 Episode_Reward/standstill: -5.9629 Episode_Reward/action_rate_l2: -0.5381 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0231 Episode_Reward/unnecessary_wheel_lift: -0.0761 Episode_Reward/rolling_leg_motion: -0.5468 Episode_Reward/obstacle_progress: +3.8202 Episode_Reward/wheel_body_intrusion: -0.2622 Episode_Reward/pair_body_intrusion: -0.2495 Curriculum/terrain_level_mean: +4.3308 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0927 Metrics/base_velocity/error_vel_yaw: 0.1361 Episode_Termination/time_out: 0.4254 Episode_Termination/fell_over: 0.0299 Episode_Termination/course_exit: 0.5075 Episode_Termination/illegal_nonwheel_contact: 0.0373 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[728] ============================================================================== Learning iteration 49550/50000 Total steps: 202,956,800 Steps per second: 13,891 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.191s Mean reward: +0.0876 Mean episode length: 1378.4 Actor loss: -9.5423 Critic loss: +0.0034 World model loss: +0.5562 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0020 Episode_Reward/track_angular_velocity: +51.9914 Episode_Reward/upright: +26.3911 Episode_Reward/leg_posture: -2.6157 Episode_Reward/standstill: -6.0549 Episode_Reward/action_rate_l2: -0.5707 Episode_Reward/joint_torques_l2: -0.1473 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0108 Episode_Reward/unnecessary_wheel_lift: -0.0806 Episode_Reward/rolling_leg_motion: -0.5767 Episode_Reward/obstacle_progress: +4.0373 Episode_Reward/wheel_body_intrusion: -0.2759 Episode_Reward/pair_body_intrusion: -0.2926 Curriculum/terrain_level_mean: +4.3328 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0923 Metrics/base_velocity/error_vel_yaw: 0.1366 Episode_Termination/time_out: 0.4965 Episode_Termination/fell_over: 0.0142 Episode_Termination/course_exit: 0.4752 Episode_Termination/illegal_nonwheel_contact: 0.0142 ============================================================================== ============================================================================== Learning iteration 49600/50000 Total steps: 203,161,600 Steps per second: 13,892 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0872 Mean episode length: 1412.9 Actor loss: -9.5217 Critic loss: +0.0051 World model loss: +0.5353 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +50.0561 Episode_Reward/track_angular_velocity: +53.2818 Episode_Reward/upright: +27.0450 Episode_Reward/leg_posture: -2.6855 Episode_Reward/standstill: -6.5461 Episode_Reward/action_rate_l2: -0.5911 Episode_Reward/joint_torques_l2: -0.1504 Episode_Reward/dof_pos_limits: -0.0008 Episode_Reward/terminated: -0.0221 Episode_Reward/unnecessary_wheel_lift: -0.1011 Episode_Reward/rolling_leg_motion: -0.6021 Episode_Reward/obstacle_progress: +4.0691 Episode_Reward/wheel_body_intrusion: -0.2761 Episode_Reward/pair_body_intrusion: -0.2761 Curriculum/terrain_level_mean: +4.3318 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0940 Metrics/base_velocity/error_vel_yaw: 0.1368 Episode_Termination/time_out: 0.5306 Episode_Termination/fell_over: 0.0272 Episode_Termination/course_exit: 0.4150 Episode_Termination/illegal_nonwheel_contact: 0.0408 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[2481] ============================================================================== Learning iteration 49650/50000 Total steps: 203,366,400 Steps per second: 13,894 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0876 Mean episode length: 1390.0 Actor loss: -9.5259 Critic loss: +0.0041 World model loss: +0.5294 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.3738 Episode_Reward/track_angular_velocity: +52.4707 Episode_Reward/upright: +26.5971 Episode_Reward/leg_posture: -2.6349 Episode_Reward/standstill: -6.0148 Episode_Reward/action_rate_l2: -0.5794 Episode_Reward/joint_torques_l2: -0.1466 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0163 Episode_Reward/unnecessary_wheel_lift: -0.0884 Episode_Reward/rolling_leg_motion: -0.6004 Episode_Reward/obstacle_progress: +3.9130 Episode_Reward/wheel_body_intrusion: -0.2547 Episode_Reward/pair_body_intrusion: -0.2774 Curriculum/terrain_level_mean: +4.3340 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0924 Metrics/base_velocity/error_vel_yaw: 0.1353 Episode_Termination/time_out: 0.4932 Episode_Termination/fell_over: 0.0137 Episode_Termination/course_exit: 0.4658 Episode_Termination/illegal_nonwheel_contact: 0.0274 ============================================================================== ============================================================================== Learning iteration 49700/50000 Total steps: 203,571,200 Steps per second: 13,896 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1369.7 Actor loss: -9.5104 Critic loss: +0.0049 World model loss: +0.5192 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.5052 Episode_Reward/track_angular_velocity: +51.6662 Episode_Reward/upright: +26.2149 Episode_Reward/leg_posture: -2.5563 Episode_Reward/standstill: -5.8055 Episode_Reward/action_rate_l2: -0.5833 Episode_Reward/joint_torques_l2: -0.1440 Episode_Reward/dof_pos_limits: -0.0006 Episode_Reward/terminated: -0.0134 Episode_Reward/unnecessary_wheel_lift: -0.0805 Episode_Reward/rolling_leg_motion: -0.5914 Episode_Reward/obstacle_progress: +3.7786 Episode_Reward/wheel_body_intrusion: -0.2524 Episode_Reward/pair_body_intrusion: -0.2961 Curriculum/terrain_level_mean: +4.3342 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0934 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4909 Episode_Termination/fell_over: 0.0061 Episode_Termination/course_exit: 0.4788 Episode_Termination/illegal_nonwheel_contact: 0.0303 ============================================================================== RECOVERY reset 1 invalid environments; first IDs=[1124] ============================================================================== Learning iteration 49750/50000 Total steps: 203,776,000 Steps per second: 13,897 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1336.2 Actor loss: -9.5371 Critic loss: +0.0034 World model loss: +0.5289 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.3793 Episode_Reward/track_angular_velocity: +50.3746 Episode_Reward/upright: +25.5695 Episode_Reward/leg_posture: -2.4333 Episode_Reward/standstill: -5.8427 Episode_Reward/action_rate_l2: -0.5730 Episode_Reward/joint_torques_l2: -0.1408 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0130 Episode_Reward/unnecessary_wheel_lift: -0.0742 Episode_Reward/rolling_leg_motion: -0.5686 Episode_Reward/obstacle_progress: +3.7213 Episode_Reward/wheel_body_intrusion: -0.2427 Episode_Reward/pair_body_intrusion: -0.2919 Curriculum/terrain_level_mean: +4.3350 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0933 Metrics/base_velocity/error_vel_yaw: 0.1371 Episode_Termination/time_out: 0.4336 Episode_Termination/fell_over: 0.0140 Episode_Termination/course_exit: 0.5315 Episode_Termination/illegal_nonwheel_contact: 0.0280 ============================================================================== Warning: EPA horizon = 24 isn't large enough. ============================================================================== Learning iteration 49800/50000 Total steps: 203,980,800 Steps per second: 13,899 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.186s Mean reward: +0.0875 Mean episode length: 1327.6 Actor loss: -9.5153 Critic loss: +0.0048 World model loss: +0.5556 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +47.0527 Episode_Reward/track_angular_velocity: +50.0746 Episode_Reward/upright: +25.4238 Episode_Reward/leg_posture: -2.4554 Episode_Reward/standstill: -5.7895 Episode_Reward/action_rate_l2: -0.5638 Episode_Reward/joint_torques_l2: -0.1405 Episode_Reward/dof_pos_limits: -0.0003 Episode_Reward/terminated: -0.0182 Episode_Reward/unnecessary_wheel_lift: -0.0905 Episode_Reward/rolling_leg_motion: -0.5694 Episode_Reward/obstacle_progress: +3.7918 Episode_Reward/wheel_body_intrusion: -0.2539 Episode_Reward/pair_body_intrusion: -0.2626 Curriculum/terrain_level_mean: +4.3354 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0935 Metrics/base_velocity/error_vel_yaw: 0.1365 Episode_Termination/time_out: 0.4072 Episode_Termination/fell_over: 0.0120 Episode_Termination/course_exit: 0.5509 Episode_Termination/illegal_nonwheel_contact: 0.0299 ============================================================================== ============================================================================== Learning iteration 49850/50000 Total steps: 204,185,600 Steps per second: 13,900 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1387.1 Actor loss: -9.5308 Critic loss: +0.0054 World model loss: +0.5315 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +49.0901 Episode_Reward/track_angular_velocity: +52.3634 Episode_Reward/upright: +26.5631 Episode_Reward/leg_posture: -2.5742 Episode_Reward/standstill: -5.7516 Episode_Reward/action_rate_l2: -0.5887 Episode_Reward/joint_torques_l2: -0.1474 Episode_Reward/dof_pos_limits: -0.0005 Episode_Reward/terminated: -0.0217 Episode_Reward/unnecessary_wheel_lift: -0.0799 Episode_Reward/rolling_leg_motion: -0.5826 Episode_Reward/obstacle_progress: +3.9813 Episode_Reward/wheel_body_intrusion: -0.2555 Episode_Reward/pair_body_intrusion: -0.3064 Curriculum/terrain_level_mean: +4.3352 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0941 Metrics/base_velocity/error_vel_yaw: 0.1355 Episode_Termination/time_out: 0.5034 Episode_Termination/fell_over: 0.0207 Episode_Termination/course_exit: 0.4414 Episode_Termination/illegal_nonwheel_contact: 0.0414 ============================================================================== ============================================================================== Learning iteration 49900/50000 Total steps: 204,390,400 Steps per second: 13,902 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0875 Mean episode length: 1315.7 Actor loss: -9.5048 Critic loss: +0.0076 World model loss: +0.5671 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.5110 Episode_Reward/track_angular_velocity: +49.6558 Episode_Reward/upright: +25.1652 Episode_Reward/leg_posture: -2.4655 Episode_Reward/standstill: -5.6186 Episode_Reward/action_rate_l2: -0.5551 Episode_Reward/joint_torques_l2: -0.1406 Episode_Reward/dof_pos_limits: -0.0011 Episode_Reward/terminated: -0.0206 Episode_Reward/unnecessary_wheel_lift: -0.0782 Episode_Reward/rolling_leg_motion: -0.5597 Episode_Reward/obstacle_progress: +3.8066 Episode_Reward/wheel_body_intrusion: -0.2579 Episode_Reward/pair_body_intrusion: -0.2917 Curriculum/terrain_level_mean: +4.3352 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0944 Metrics/base_velocity/error_vel_yaw: 0.1359 Episode_Termination/time_out: 0.4015 Episode_Termination/fell_over: 0.0365 Episode_Termination/course_exit: 0.5401 Episode_Termination/illegal_nonwheel_contact: 0.0365 ============================================================================== ============================================================================== Learning iteration 49950/50000 Total steps: 204,595,200 Steps per second: 13,904 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0878 Mean episode length: 1378.0 Actor loss: -9.5147 Critic loss: +0.0050 World model loss: +0.5321 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +48.8101 Episode_Reward/track_angular_velocity: +52.0469 Episode_Reward/upright: +26.3670 Episode_Reward/leg_posture: -2.5521 Episode_Reward/standstill: -5.7972 Episode_Reward/action_rate_l2: -0.5890 Episode_Reward/joint_torques_l2: -0.1461 Episode_Reward/dof_pos_limits: -0.0009 Episode_Reward/terminated: -0.0188 Episode_Reward/unnecessary_wheel_lift: -0.0759 Episode_Reward/rolling_leg_motion: -0.5804 Episode_Reward/obstacle_progress: +4.0427 Episode_Reward/wheel_body_intrusion: -0.2530 Episode_Reward/pair_body_intrusion: -0.2815 Curriculum/terrain_level_mean: +4.3364 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0938 Metrics/base_velocity/error_vel_yaw: 0.1350 Episode_Termination/time_out: 0.4937 Episode_Termination/fell_over: 0.0190 Episode_Termination/course_exit: 0.4620 Episode_Termination/illegal_nonwheel_contact: 0.0253 ============================================================================== ============================================================================== Learning iteration 50000/50000 Total steps: 204,800,000 Steps per second: 13,905 Iteration time: 0.295s Collection time: 0.076s Learning time: 0.185s Mean reward: +0.0877 Mean episode length: 1311.9 Actor loss: -9.5213 Critic loss: +0.0061 World model loss: +0.5613 Mean action noise: 0.150 Skipped updates: 0 ------------------------------------------------------------------------------ Episode_Reward/track_linear_velocity: +46.4991 Episode_Reward/track_angular_velocity: +49.4780 Episode_Reward/upright: +25.0778 Episode_Reward/leg_posture: -2.4390 Episode_Reward/standstill: -5.5298 Episode_Reward/action_rate_l2: -0.5605 Episode_Reward/joint_torques_l2: -0.1406 Episode_Reward/dof_pos_limits: -0.0007 Episode_Reward/terminated: -0.0307 Episode_Reward/unnecessary_wheel_lift: -0.0838 Episode_Reward/rolling_leg_motion: -0.5458 Episode_Reward/obstacle_progress: +3.8305 Episode_Reward/wheel_body_intrusion: -0.2444 Episode_Reward/pair_body_intrusion: -0.2655 Curriculum/terrain_level_mean: +4.3359 Command/command_stage: 1 Command/command_ranges: [[-0.5000, +1.0000], [-0.4000, +0.4000], [-0.7000, +0.7000]] Metrics/base_velocity/error_vel_xy: 0.0937 Metrics/base_velocity/error_vel_yaw: 0.1364 Episode_Termination/time_out: 0.4557 Episode_Termination/fell_over: 0.0380 Episode_Termination/course_exit: 0.4684 Episode_Termination/illegal_nonwheel_contact: 0.0633 ============================================================================== +---------------------------------+ | Base Environment | +------------------------+--------+ | Property | Value | +------------------------+--------+ | Number of environments | 1 | | Environment device | cuda:0 | | Environment seed | None | | Physics step-size | 0.005 | | Environment step-size | 0.02 | +------------------------+--------+ [INFO] contains 1 active terms. +-------------------------------------+ | Active Event Terms in Mode: 'reset' | +--------+----------------------------+ | Index | Name | +--------+----------------------------+ | 0 | reset_base | | 1 | reset_robot_joints | +--------+----------------------------+ [INFO] contains 1 active terms. +--------------------------------------+ | Active Command Terms | +-------+-------+----------------------+ | Index | Name | Type | +-------+-------+----------------------+ | 0 | twist | MixedVelocityCommand | +-------+-------+----------------------+ [INFO] contains 2 active terms. +---------------------------------+ | Active Action Terms (shape: 16) | +-------+-----------+-------------+ | Index | Name | Dimension | +-------+-----------+-------------+ | 0 | joint_pos | 12 | | 1 | wheel_vel | 4 | +-------+-----------+-------------+ [INFO] contains 1 groups. +------------------------------------------------------------+ | Active Observation Terms in Group: 'actor' (shape: (265,)) | +--------+-------------------------+-------------------------+ | Index | Name | Shape | +--------+-------------------------+-------------------------+ | 0 | base_ang_vel | (15,) ← 5×(3,) | | 1 | projected_gravity | (15,) ← 5×(3,) | | 2 | joint_pos | (60,) ← 5×(12,) | | 3 | joint_vel | (60,) ← 5×(12,) | | 4 | wheel_vel | (20,) ← 5×(4,) | | 5 | actions | (80,) ← 5×(16,) | | 6 | command | (15,) ← 5×(3,) | +--------+-------------------------+-------------------------+ [INFO] contains 4 active terms. +---------------------------------------------+ | Active Termination Terms | +-------+--------------------------+----------+ | Index | Name | Time Out | +-------+--------------------------+----------+ | 0 | time_out | True | | 1 | fell_over | False | | 2 | course_exit | True | | 3 | illegal_nonwheel_contact | False | +-------+--------------------------+----------+ [INFO] contains 0 active terms. +-----------------------+ | Active Reward Terms | +-------+------+--------+ | Index | Name | Weight | +-------+------+--------+ +-------+------+--------+ [INFO] (inactive) [INFO] (inactive) [INFO] (inactive) S10 video saved: logs/s10_terrain_v7_50k_20261101_225518/videos/iteration_050000.mp4 S10 training complete: logs/s10_terrain_v7_50k_20261101_225518