Observation builders for MuJoCo sim2sim deployments.
Functions:
| Name | Description |
|---|---|
build_mujoco_state |
Extract the unbatched controller state exposed by the LAV2 MJCF. |
trajectory_preview_command |
Sample a training-compatible trajectory preview in yaw-body frame. |
build_trajectory_observation |
Build the fixed aerial/ground trajectory-policy observation. |
landing_target_command |
Return the fixed four-channel landing target in yaw-body frame. |
build_landing_observation |
Build the fixed landing-policy observation in training term order. |
build_default_flight_observation |
Build the legacy base-task observation used by current MuJoCo play scripts. |
build_simpleflight_observation |
Build the Isaac Lab SimpleFlight observation. |
make_simpleflight_observation_extractor |
Return an observation extractor matching the SimpleFlight task interface. |
build_mujoco_state(d: MjData) -> ControllerState
Extract the unbatched controller state exposed by the LAV2 MJCF.
trajectory_preview_command(trajectory: BaseTrajectory, state: ControllerState, trajectory_time: float, *, command_mode: str = 'tgt_pos', future_steps: int = TRAJECTORY_FUTURE_STEPS, preview_dt: float = TRAJECTORY_PREVIEW_DT) -> tuple[np.ndarray, ControllerTarget]
Sample a training-compatible trajectory preview in yaw-body frame.
build_trajectory_observation(state: ControllerState, preview_command: ndarray, last_action: ndarray) -> np.ndarray
Build the fixed aerial/ground trajectory-policy observation.
landing_target_command(state: ControllerState, target: ControllerTarget) -> np.ndarray
Return the fixed four-channel landing target in yaw-body frame.
build_landing_observation(state: ControllerState, target_command: ndarray, last_action: ndarray) -> np.ndarray
Build the fixed landing-policy observation in training term order.
build_default_flight_observation(d: MjData, target: ControllerTarget, last_action: ndarray, command_mode: str, output_frame: str = 'yaw_body') -> np.ndarray
Build the legacy base-task observation used by current MuJoCo play scripts.
build_simpleflight_observation(d: MjData, trajectory: BaseTrajectory | None, target: ControllerTarget, cfg: MuJoCoPlayConfig) -> np.ndarray
Build the Isaac Lab SimpleFlight observation.
make_simpleflight_observation_extractor(cfg: MuJoCoPlayConfig, trajectory: BaseTrajectory | None)
Return an observation extractor matching the SimpleFlight task interface.