Parameterized base environment configuration for mjlab aerial tasks.
This module provides robot-agnostic factory functions for building
ManagerBasedRlEnvCfg instances. Concrete robot platforms (e.g. LAV2)
subclass or wrap these with platform-specific scene, viewer, and actuator
defaults in config/<robot>/env_cfgs.py.
Functions:
| Name | Description |
|---|---|
create_scene_cfg |
Build a scene config for the given robot entity. |
create_action_cfg |
Build the action-manager config. |
create_observation_cfg |
Build the observation-manager config (robot-agnostic terms). |
create_event_cfg |
Build the event-manager config (robot-agnostic domain-randomisation). |
create_command_cfg |
Build the command-manager config for pose-command tracking. |
create_reward_cfg |
Build the reward-manager config (robot-agnostic). |
create_termination_cfg |
Build the termination-manager config (robot-agnostic). |
create_viewer_config |
Build a viewer config anchored to an asset body. |
create_simulation_cfg |
Build the simulation config. |
create_env_cfg |
Build a complete |
create_scene_cfg(robot_cfg: EntityCfg, *, num_envs: int = 1, extent: float = 2.0) -> SceneCfg
Build a scene config for the given robot entity.
create_action_cfg(actuator_names: tuple[str, ...] | None = None) -> dict[str, ActionTermCfg]
Build the action-manager config.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
|
tuple[str, ...] | None
|
Override the default wrench-actuator list. When
None the |
None
|
create_observation_cfg() -> dict[str, ObservationGroupCfg]
Build the observation-manager config (robot-agnostic terms).
create_event_cfg() -> dict[str, EventTermCfg]
Build the event-manager config (robot-agnostic domain-randomisation).
create_command_cfg(entity_name: str = 'robot') -> dict[str, CommandTermCfg]
Build the command-manager config for pose-command tracking.
create_reward_cfg() -> dict[str, RewardTermCfg]
Build the reward-manager config (robot-agnostic).
create_termination_cfg() -> dict[str, TerminationTermCfg]
Build the termination-manager config (robot-agnostic).
create_viewer_config(entity_name: str = 'robot', body_name: str = 'base_link') -> ViewerConfig
Build a viewer config anchored to an asset body.
create_simulation_cfg(timestep: float = 0.01, iterations: int = 2) -> SimulationCfg
Build the simulation config.
create_env_cfg(*, robot_cfg: EntityCfg, entity_name: str = 'robot', body_name: str = 'base_link', actuator_names: tuple[str, ...] | None = None, scene_num_envs: int = 1, scene_extent: float = 2.0, sim_timestep: float = 0.01, sim_iterations: int = 2, decimation: int = 2, episode_length_s: float = 10.0) -> ManagerBasedRlEnvCfg
Build a complete ManagerBasedRlEnvCfg for an aerial base task.
Robot-platform specifics (scene entity, viewer body name, actuator
list) are injected via parameters so the same builder works for LAV2,
ATMO, or any other multirotor supported by FlightControlActionCfg.