add randomize_motor_zero_offset to align with go2_rl_gym.

README.md

@@ -161,7 +161,6 @@ xml_path: "{ROOT_DIR}/resources/go2/your-custom-scene.xml"
 
 - Different terrain composition
 - Different tracking reward formulation (fixed sigma vs. dynamic sigma)
-- Lack domain_rand: randomize_motor_zero_offset
 - Lack domain_rand: randomize_motor_strength
 
 ---

source/robot_lab/robot_lab/tasks/go2/__init__.py

@@ -19,6 +19,7 @@ from isaaclab_tasks.utils import import_packages
 ##
 gym.register(
     id="RobotLab-Go2-v0",
+    # entry_point="isaaclab.envs:ManagerBasedRLEnv",
     entry_point="robot_lab.tasks.go2.env.go2_env:ActionDelayGo2Env",
     disable_env_checker=True,
     kwargs={

source/robot_lab/robot_lab/tasks/go2/env_cfg.py

@@ -346,6 +346,14 @@ class EventCfg:
             "distribution": "uniform",
         },
     )
+    randomize_motor_zero_offset = EventTerm(
+        func=mdp.randomize_action_joint_pos_offset,
+        mode="reset",
+        params={
+            "action_term_name": "joint_pos",
+            "offset_range": (-0.035, 0.035),
+        },
+    )
     randomize_push_robot = EventTerm(
         func=mdp.push_by_setting_velocity,
         mode="interval",

source/robot_lab/robot_lab/tasks/go2/mdp/events.py

@@ -267,3 +267,40 @@ def reset_root_state_uniform(
         # set into the physics simulation
         asset.write_root_pose_to_sim(torch.cat([positions, orientations], dim=-1), env_ids=non_pit_env_ids)
         asset.write_root_velocity_to_sim(velocities, env_ids=non_pit_env_ids)
+
+def randomize_action_joint_pos_offset(
+    env: ManagerBasedEnv,
+    env_ids: torch.Tensor | None,
+    action_term_name: str,
+    offset_range: tuple[float, float],
+):
+    """Randomize the motor zero-offset on a joint-position action term."""
+    if env_ids is None:
+        env_ids = torch.arange(env.scene.num_envs, device=env.device)
+    elif not isinstance(env_ids, torch.Tensor):
+        env_ids = torch.as_tensor(env_ids, dtype=torch.long, device=env.device)
+    else:
+        env_ids = env_ids.to(device=env.device, dtype=torch.long)
+    if len(env_ids) == 0:
+        return
+
+    action_term = env.action_manager.get_term(action_term_name)
+    if not hasattr(action_term, "_offset") or not isinstance(action_term._offset, torch.Tensor):
+        raise TypeError(
+            f"Action term '{action_term_name}' does not expose a tensor '_offset', "
+            "so it cannot be used for motor zero-offset randomization."
+        )
+
+    cache_name = f"_default_action_offset_{action_term_name}"
+    default_offset = getattr(env, cache_name, None)
+    if default_offset is None or default_offset.shape != action_term._offset.shape:
+        default_offset = action_term._offset.clone()
+        setattr(env, cache_name, default_offset)
+
+    offset_noise = math_utils.sample_uniform(
+        offset_range[0],
+        offset_range[1],
+        (len(env_ids), action_term._offset.shape[1]),
+        device=env.device,
+    )
+    action_term._offset[env_ids] = default_offset[env_ids] + offset_noise