函数：
| 名称 | 描述 |
|---|---|
xy_goal_tanh |
Reward small horizontal error with a height-scheduled tanh tolerance. |
descent_velocity_exp |
Reward a constant descent speed until first ground contact. |
touchdown_contact_reward |
Reward contact only when touchdown is near the landing point and slow. |
touchdown_action_target_l2 |
Penalize action error from the motor-off target after touchdown. |
has_contact |
Return true for envs where any selected body is currently in contact. |
xy_goal_tanh(env: ManagerBasedRLEnv, command_name: str, std: float, height_scale: float = 2.0, min_height: float = 0.05, max_height: float = 1.0) -> torch.Tensor
Reward small horizontal error with a height-scheduled tanh tolerance.
descent_velocity_exp(env: ManagerBasedRLEnv, target_vz: float, std: float, command_name: str, sensor_cfg: SceneEntityCfg | None = None, asset_cfg: SceneEntityCfg = SceneEntityCfg('robot')) -> torch.Tensor
Reward a constant descent speed until first ground contact.
touchdown_contact_reward(env: ManagerBasedRLEnv, command_name: str, xy_radius: float, max_speed: float, sensor_cfg: SceneEntityCfg, asset_cfg: SceneEntityCfg = SceneEntityCfg('robot')) -> torch.Tensor
Reward contact only when touchdown is near the landing point and slow.
touchdown_action_target_l2(env: ManagerBasedRLEnv, sensor_cfg: SceneEntityCfg, collective_target: float = -1.0) -> torch.Tensor
Penalize action error from the motor-off target after touchdown.
This follows Isaac Lab's squared-L2 action penalty, but the collective
channel targets -1 because the LAV2 flight mapping converts -1 to
zero thrust and 0 to half maximum collective thrust. Body-rate channels
continue to target zero.
has_contact(env: ManagerBasedRLEnv, sensor_cfg: SceneEntityCfg) -> torch.Tensor
Return true for envs where any selected body is currently in contact.