Commit Graph

  • 3963773b6f Sample symmetric DreamWaQ speed commands main 8x54zj-m 2026-07-22 20:08:07 +08:00
  • 3fa2849123 Sample forward DreamWaQ speed commands 8x54zj-m 2026-07-22 20:03:02 +08:00
  • 509b3d635c Cap DreamWaQ adaptive learning rate 8x54zj-m 2026-07-22 19:58:19 +08:00
  • 944e91e28a Require consecutive failures for terrain downgrade 8x54zj-m 2026-07-22 19:48:48 +08:00
  • d70b084318 Revert "Limit DreamWaQ commands to stand or walk" 8x54zj-m 2026-07-22 19:42:40 +08:00
  • 6d6a805091 Limit DreamWaQ commands to stand or walk 8x54zj-m 2026-07-22 19:39:13 +08:00
  • d22292d726 Revert "Use integrated progress for terrain curriculum" 8x54zj-m 2026-07-22 19:37:26 +08:00
  • 7358b31de1 Use integrated progress for terrain curriculum 8x54zj-m 2026-07-22 19:10:48 +08:00
  • 913d5061ff Fix terrain curriculum traversal and hfield roughness 8x54zj-m 2026-07-22 16:46:32 +08:00
  • a0e8501d11 Add independent DreamWaQ stair capability benchmark 8x54zj-m 2026-07-22 16:31:09 +08:00
  • 82f4997deb Align DreamWaQ Go1 environment rewards and initialization 8x54zj-m 2026-07-22 14:08:10 +08:00
  • 0555776c11 Align DreamWaQ PPO rollout and policy dynamics 8x54zj-m 2026-07-22 13:59:16 +08:00
  • 57faf8880b Bootstrap RSLRL rewards on environment timeouts 8x54zj-m 2026-07-22 13:30:25 +08:00
  • 1072565d97 Fix action smoothing history and viewer timing control 8x54zj-m 2026-07-22 13:25:17 +08:00
  • 1a84cf69ab Reset invalid DreamWaQ physics states safely 8x54zj-m 2026-07-22 04:30:13 +08:00
  • 660393a9ee Add closed-loop DreamWaQ stair validation 8x54zj-m 2026-07-22 02:36:36 +08:00
  • 9cdf59a99d Add DreamWaQ ONNX sim2sim pipeline 8x54zj-m 2026-07-22 02:28:03 +08:00
  • 00cb6f9857 Tune DreamWaQ gait rewards 8x54zj-m 2026-07-22 02:22:04 +08:00
  • 3648551043 Backup DreamWaQ rslrl stability fixes 8x54zj-m 2026-07-22 02:17:01 +08:00
  • c664e7422f 6level,但是表现一般 8x54zj-m 2026-07-02 10:53:17 +08:00
  • 482dabb893 微调参数 8x54zj-m 2026-07-01 21:02:40 +08:00
  • 63b235d186 fix: MuJoCo sim2sim default_angles match MotrixSim training 8x54zj-m 2026-07-01 20:55:16 +08:00
  • 2bd4c05e1b fix: fixed schedule, init_noise 0.5, AdaBoot cap 0.9 (keep vel head gradient) 8x54zj-m 2026-07-01 16:56:07 +08:00
  • a3562fbab1 fix: storage dones/rewards shape squeeze, NaN guard in AdaBoot 8x54zj-m 2026-07-01 16:02:21 +08:00
  • be193199e5 feat: AdaBoot — paper CV of episodic reward (not velocity error) 8x54zj-m 2026-07-01 15:57:00 +08:00
  • 5cbc8e2ca4 feat: AdaBoot — adaptive velocity bootstrapping (paper Section II-C) 8x54zj-m 2026-07-01 15:44:30 +08:00
  • 308f2090a5 fix: VAE predicts next frame (paper: LVAE = MSE(õ_{t+1}, o_{t+1})), history excludes current obs 8x54zj-m 2026-07-01 15:39:56 +08:00
  • 3038aa7ac7 fix: critic called once (was 3x), adaptive KL schedule restored, upgrade 3m 8x54zj-m 2026-07-01 15:05:50 +08:00
  • ef3ccb2e1a fix: NaN protection — dof nan_to_num, physics reset on crash 8x54zj-m 2026-07-01 10:51:36 +08:00
  • 35a6f66511 fix: NaN chain break — rewards, std, obs, physics all protected 8x54zj-m 2026-07-01 10:31:47 +08:00
  • 1eede03ef0 fix: dual-shuffle bug in terrain_origins — robots now spawn on correct terrain columns 8x54zj-m 2026-06-30 22:16:29 +08:00
  • cb62f35ef6 fix: boundary termination, physics crash handler, action_scale 0.25, play script update 8x54zj-m 2026-06-30 19:03:57 +08:00
  • 7affccc695 fix: noise scales *obs_scale (match upstream), noise_joint_angle 0.03→0.01 8x54zj-m 2026-06-30 17:30:24 +08:00
  • 964b409aa1 fix: action_scale 0.05→0.25 (upstream value, 5x difference) 8x54zj-m 2026-06-30 17:29:16 +08:00
  • a5e5d5029e fix: dt in state.reward, feet_air_time first_contact timing 8x54zj-m 2026-06-30 16:42:48 +08:00
  • 361502a8a5 fix: terrain start level 0, clear ep_report after consume 8x54zj-m 2026-06-30 15:43:02 +08:00
  • 6348cf72bc fix: squeeze old_log_prob and advantages dims (match upstream) 8x54zj-m 2026-06-30 15:39:03 +08:00
  • aa17ea853e fix: VAE loss /num_mini_batches, KL use raw sum (align with upstream) 8x54zj-m 2026-06-30 15:38:26 +08:00
  • b0f1da4596 fix: clamp std before distribution, lower init_noise to 0.5, NaN guard 8x54zj-m 2026-06-30 15:35:58 +08:00
  • f2a8e0e2ff fix: remove broken feet_air_time override, use base class 8x54zj-m 2026-06-30 15:03:52 +08:00
  • 3fd2793f28 fix: register dreamwaq env, disable missing stair envs 8x54zj-m 2026-06-30 14:56:01 +08:00
  • 422421d263 feat: DreamWaQ full replication — env, terrain, CENet, PPO 8x54zj-m 2026-06-30 14:53:12 +08:00
  • c569f4d6d9 Merge pull request #5 from Motphys/docs/add-motrixlab-citation motphys-developers 2026-05-25 16:46:11 +08:00
  • e1421d1055 chore: release v0.3.0 motphys-developers 2026-04-02 03:45:10 +00:00
  • c84d382b8c chore: release v0.2.0.post1 motphys-developers 2026-02-10 16:53:35 +08:00
  • b568ac5600 chore: release v0.2.0 motphys-developers 2026-02-10 08:08:11 +00:00
  • dbfa9e31fa add go2 training jyf23 2026-01-02 21:10:34 +08:00
  • 62011bb24f chore: release v0.1.0 motphys-developers 2026-01-04 04:43:04 +00:00
  • 13cfbce9a7 chore: release v0.0.2 motphys-developers 2025-11-21 08:51:49 +00:00
  • 5133830b5a chore: release v0.0.1 motphys-developers 2025-11-20 08:57:48 +00:00