v1.0.4; slope_thre: 0.7->2.0

README.md

@@ -67,7 +67,7 @@ The trained model above was evaluated using the [RoboGauge](https://github.com/w
 | [HIM](https://github.com/InternRobotics/HIMLoco) | 0.5379 | 0.5453 | 0.6476 | 0.6050 | 6.19 | [ckpt](https://drive.google.com/file/d/1remJbGoTorqnArsz8Z1ewY4TVobss4Fb/view?usp=drive_link) |
 | [DreamWaQ](https://arxiv.org/abs/2301.10602) | 0.5054 | 0.5105 | 0.6149 | 0.5730 | 5.74 | [ckpt](https://drive.google.com/file/d/19BEBeiQqjHcPgGrN3AX6D7Yefs_8eswL/view?usp=drive_link) |
 
-> In the downloaded ckpt files, `*.pt` is used for [Python deployment](#41-python-deployment), and `*.onnx` is used for [C++ deployment](#42-c-deployment). The models above were all trained with self-collision disabled. In later tests, we found that enabling self-collision can also achieve strong results; see [go2_moe_cts_self_0.6669 - ckpt](https://drive.google.com/drive/folders/1znytqHNtDiZM5J4vaBd-EuM81l91D6s5?usp=drive_link).
+> In the downloaded ckpt files, `*.pt` is used for [Python deployment](#41-python-deployment), and `*.onnx` is used for [C++ deployment](#42-c-deployment). The models above were all trained with self-collision disabled. In later tests, we found that enabling self-collision can also achieve strong results; see [go2_moe_cts_164k_0.6715 - exported](https://drive.google.com/drive/folders/1w8ctwb77PE7wDnlC-XYY1CSh39e4IQmT?usp=drive_link) with [complete model weights - model_164000.pt](https://drive.google.com/file/d/1mwQywpM6UpzZWzHOD_MMxWqgFKLr5UZw/view?usp=drive_link).
 
 ### 2. Play
 

README_zh.md

@@ -67,7 +67,7 @@ python legged_gym/scripts/train.py --task=xxx --headless
 | [HIM](https://github.com/InternRobotics/HIMLoco) | 0.5379 | 0.5453 | 0.6476 | 0.6050 | 6.19 | [ckpt](https://drive.google.com/file/d/1remJbGoTorqnArsz8Z1ewY4TVobss4Fb/view?usp=drive_link) |
 | [DreamWaQ](https://arxiv.org/abs/2301.10602) | 0.5054 | 0.5105 | 0.6149 | 0.5730 | 5.74 | [ckpt](https://drive.google.com/file/d/19BEBeiQqjHcPgGrN3AX6D7Yefs_8eswL/view?usp=drive_link) |
 
-> 下载的 ckpt 中，`*.pt` 用于[Python 实物部署](#41-python实物部署)，`*.onnx` 用于[C++ 实物部署](#42-c实物部署)。上述模型均在关闭自碰撞的设置下训练；后续测试发现，开启自碰撞也能取得不错效果，参考 [go2_moe_cts_self_0.6669 - ckpt](https://drive.google.com/drive/folders/1znytqHNtDiZM5J4vaBd-EuM81l91D6s5?usp=drive_link)。
+> 下载的 ckpt 中，`*.pt` 用于[Python 实物部署](#41-python实物部署)，`*.onnx` 用于[C++ 实物部署](#42-c实物部署)。上述模型均在关闭自碰撞的设置下训练；后续测试发现，开启自碰撞也能取得不错效果，参考 [go2_moe_cts_164k_0.6715 - exported](https://drive.google.com/drive/folders/1w8ctwb77PE7wDnlC-XYY1CSh39e4IQmT?usp=drive_link)以及其[完整模型权重 - model_164000.pt](https://drive.google.com/file/d/1mwQywpM6UpzZWzHOD_MMxWqgFKLr5UZw/view?usp=drive_link)。
 
 ---
 

UPDATE.md

@@ -1,3 +1,6 @@
+# 20260419
+## v1.0.4
+1. 地形中`slope_threshold: 0.75 -> 1.5`，wave和rough_slope表现更好
 # 20260403
 ## v1.0.3
 1. 修复last_last_action重置问题，修复resample_command在计算奖励前的问题

legged_gym/envs/base/legged_robot_config.py

@@ -37,7 +37,7 @@ class LeggedRobotCfg(BaseConfig):
         # [wave, slope, rough slope, stairs up, stairs down, obstacles, stepping stones, gap, flat]
         terrain_proportions = [0.1, 0.1, 0.1, 0.2, 0.2, 0.1, 0.1, 0.1, 0.0]
         # trimesh only:
-        slope_treshold = 0.75 # slopes above this threshold will be corrected to vertical surfaces
+        slope_threshold = 0.75 # slopes above this threshold will be corrected to vertical surfaces
         move_down_by_accumulated_xy_command = False # move down the terrain curriculum based on accumulated xy command distance instead of absolute distance
 
     class commands:

legged_gym/envs/go2/go2_config.py

@@ -93,6 +93,7 @@ class GO2Cfg(LeggedRobotCfg):
         # terrain_proportions = [0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]
         # terrain_proportions = [0.3, 0.3, 0.3, 0.0, 0.0, 0.0, 0.0, 0.0, 0.1]
         # terrain_proportions = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0]
+        slope_threshold = 1.5  # higher value means more steep slopes, for better wave and rough slope generation
         move_down_by_accumulated_xy_command = True # move down the terrain curriculum based on accumulated xy command distance instead of absolute distance
         
     class commands(LeggedRobotCfg.commands):

legged_gym/utils/terrain.py

@@ -46,7 +46,7 @@ class Terrain:
             self.vertices, self.triangles = terrain_utils.convert_heightfield_to_trimesh(   self.height_field_raw,
                                                                                             self.cfg.horizontal_scale,
                                                                                             self.cfg.vertical_scale,
-                                                                                            self.cfg.slope_treshold)
+                                                                                            self.cfg.slope_threshold)
     
     def randomized_terrain(self):
         for k in range(self.cfg.num_sub_terrains):