actions

类：

名称	描述
FlightControlActionManager	Base class for Genesis flight action managers.

FlightControlActionManager

FlightControlActionManager(env: GenesisEnv, delay_step: int = 0, params: VehicleParams = VehicleParams(), gains: dict | None = None, limits: dict | None = None, mapping: FlightMappingConfig = FlightMappingConfig(), entity_attr: str = 'robot', rotor_link_names: tuple[str, ...] = ('rotor0', 'rotor1', 'rotor2', 'rotor3'), control_mode: str = 'cmd_motor_speeds', controller: str = 'FlightController', randomize_gains: bool | dict = False)

Bases: BaseActionManager

Base class for Genesis flight action managers.

Initialize the Genesis flight action manager.

参数：

名称	类型	描述	默认
env	GenesisEnv	The environment that owns the controlled robot entity.	必需
delay_step	int	Number of control steps to delay incoming actions.	0
params	VehicleParams	Vehicle physical parameters.	VehicleParams()
gains	dict | None	Optional flight-controller gain overrides.	None
limits	dict | None	Optional flight-controller command-limit overrides.	None
mapping	FlightMappingConfig	Action-mapping overrides resolved against params.	FlightMappingConfig()
entity_attr	str	Environment attribute name for the robot entity.	'robot'
rotor_link_names	tuple[str, ...]	Ordered rotor link names matching controller/mixer order.	('rotor0', 'rotor1', 'rotor2', 'rotor3')
control_mode	str	Flight control mode for interpreting the incoming action.	'cmd_motor_speeds'
controller	str	Controller class name. Options: 'FlightController' (PID), 'GeoControl'.	'FlightController'
randomize_gains	bool | dict	If True, gains are randomized on reset() via gains_to_randomization_ranges(). Pass a dict of min/max ranges to use custom bounds instead.	False

方法：

名称	描述
step	Process the incoming actions (analogous to Isaac Lab process_actions).
send_actions_to_simulation	Apply the processed actions to the simulation (Isaac Lab apply_actions).
reset	Reset environments.
get_actions	Get the current actions for the environments.

属性：

名称	类型	描述
num_actions	int	The total number of actions.
action_space	tuple[float, float]	If using the default action handler, the action space is [-1, 1].
actions	Tensor	The actions for for the current step.
raw_actions	Tensor	The actions received from the policy, before being converted.

num_actions property

num_actions: int

The total number of actions.

action_space property

action_space: tuple[float, float]

If using the default action handler, the action space is [-1, 1].

actions property

actions: Tensor

The actions for for the current step.

raw_actions property

raw_actions: Tensor

The actions received from the policy, before being converted.

step

step(actions: Tensor) -> torch.Tensor

Process the incoming actions (analogous to Isaac Lab process_actions).

Only the raw and clamped policy actions are stored here; the flight controller, mixer, rotor dynamics and body-force application all run in :meth:send_actions_to_simulation (the apply phase). Keeping the controller in the apply phase lets it read fresh robot state on each apply tick, mirroring Isaac Lab's process_actions/apply_actions split.

send_actions_to_simulation

send_actions_to_simulation(processed_actions: Tensor) -> torch.Tensor

Apply the processed actions to the simulation (Isaac Lab apply_actions).

genesis_forge>=0.5 splits action handling into step (process) and send_actions_to_simulation (apply), which runs at a different cadence than processing. This reads fresh robot state, runs the flight controller/mixer to obtain the rotor RPM command, steps the rotor dynamics, and writes the resulting body forces/torques to the solver.

reset

reset(envs_idx: list[int] | None)

Reset environments.

get_actions

get_actions() -> torch.Tensor

Get the current actions for the environments.