Independent three-policy AeroCar runtime for MuJoCo sim2sim.
Classes:
| Name | Description |
|---|---|
AeroCarSimState |
Policy phase owned entirely by the MuJoCo sim2sim runtime. |
AeroCarMuJoCoConfig |
Configuration for independent AeroCar MuJoCo policy playback. |
AeroCarMuJoCoRunner |
Run aerial trajectory, landing, and ground policies in MuJoCo. |
Bases: str, Enum
Policy phase owned entirely by the MuJoCo sim2sim runtime.
dataclass
AeroCarMuJoCoConfig(aerial_model_path: str = '', landing_model_path: str = '', ground_model_path: str = '', device: str = 'cpu', scene_path: str = str(LAV2_ASSETS.scene_path), initial_state: AeroCarSimState = AeroCarSimState.AERIAL_TRAJ, initial_altitude: float = 1.0, aerial_trajectory: str = 'helix', ground_trajectory: str = 'rectangle', trajectory_time_offset: float = 0.0, ground_z: float = 0.0, landing_target_height: float = 0.05, touchdown_max_vertical_speed: float = 0.25, touchdown_stable_s: float = 0.4, decimation: int = 2, ground_actuator: GroundDriveActuatorConfig = (lambda: GroundDriveActuatorConfig(yaw_allocation_scale=2.1, left_delay_s=0.01, right_delay_s=0.03))(), log_output_folder: str = 'results', save_log: bool = True, plot_log: bool = True, params: TrackVehicleParams = (lambda: copy.deepcopy(LAV2_ASSETS.vehicle_params))(), flight_mapping: FlightMappingConfig = FlightMappingConfig(), ground_mapping: DifferentialDriveMappingConfig = DifferentialDriveMappingConfig())
Configuration for independent AeroCar MuJoCo policy playback.
Methods:
| Name | Description |
|---|---|
__post_init__ |
Validate settings used directly by the MuJoCo loop. |
__post_init__() -> None
Validate settings used directly by the MuJoCo loop.
AeroCarMuJoCoRunner(cfg: AeroCarMuJoCoConfig, *, policy_runners: dict[str, Any] | None = None)
Run aerial trajectory, landing, and ground policies in MuJoCo.
Load policies and initialize MuJoCo-specific controllers.
Methods:
| Name | Description |
|---|---|
request_landing |
Request the normal aerial-to-landing transition. |
force_state |
Force a policy phase for isolated interactive playback. |
reset |
Reset controllers and reactivate the configured initial phase. |
step |
Advance the policy FSM and produce a MuJoCo actuator command. |
control_callback |
Run policy inference at decimated rate and apply latched controls. |
key_callback |
Handle interactive phase selection in the MuJoCo viewer. |
load_callback |
Load the LAV2 scene and register the MuJoCo control callback. |
run |
Launch the interactive MuJoCo AeroCar playback session. |
request_landing() -> bool
Request the normal aerial-to-landing transition.
force_state(new_state: AeroCarSimState, now_sec: float) -> bool
Force a policy phase for isolated interactive playback.
reset() -> None
Reset controllers and reactivate the configured initial phase.
step(m: MjModel, d: MjData) -> dict
Advance the policy FSM and produce a MuJoCo actuator command.
control_callback(m: MjModel, d: MjData) -> None
Run policy inference at decimated rate and apply latched controls.
key_callback(keycode: int) -> None
Handle interactive phase selection in the MuJoCo viewer.
load_callback() -> tuple[mujoco.MjModel, mujoco.MjData]
Load the LAV2 scene and register the MuJoCo control callback.
run() -> None
Launch the interactive MuJoCo AeroCar playback session.