4090工作站代码

scripts/rsl_rl/train.py

@@ -184,6 +184,22 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # wrap around environment for rsl-rl
     env = RslRlVecEnvWrapper(env, clip_actions=agent_cfg.clip_actions)
 
+    # # 添加pdb断点来打印left_arm_ee
+    # import pdb
+    # original_step = env.step
+    # def step_with_debug(action):
+    #     result = original_step(action)
+    #     # 访问底层环境的action_manager
+    #     unwrapped_env = env.unwrapped
+    #     if hasattr(unwrapped_env, 'action_manager'):
+    #         left_arm_ee_term = unwrapped_env.action_manager.get_term("left_arm_ee")
+    #         if left_arm_ee_term is not None:
+    #             print(f"left_arm_ee raw_actions: {left_arm_ee_term.raw_actions}")
+    #             print(f"left_arm_ee processed_actions: {left_arm_ee_term.processed_actions}")
+    #             pdb.set_trace()  # 在这里设置断点
+    #     return result
+    # env.step = step_with_debug
+
     # create runner from rsl-rl
     if agent_cfg.class_name == "OnPolicyRunner":
         runner = OnPolicyRunner(env, agent_cfg.to_dict(), log_dir=log_dir, device=agent_cfg.device)