mani_skill.agents.controllers.pd_joint_pos

Classes

MimicConfig	dict() -> new empty dictionary
PDJointPosController	Base class for controllers.
PDJointPosControllerConfig
PDJointPosMimicController	Base class for controllers.
PDJointPosMimicControllerConfig	PD Joint Position mimic controller configuration. This kind of controller is used to emuluate a mimic joint. For some simulation backends mimic joints are not

Module Contents

class mani_skill.agents.controllers.pd_joint_pos.MimicConfig

Bases: TypedDict

dict() -> new empty dictionary dict(mapping) -> new dictionary initialized from a mapping object’s

(key, value) pairs

dict(iterable) -> new dictionary initialized as if via:

d = {} for k, v in iterable:

d[k] = v

dict(**kwargs) -> new dictionary initialized with the name=value pairs

in the keyword argument list. For example: dict(one=1, two=2)

joint: str

multiplier: NotRequired[float]

offset: NotRequired[float]

class mani_skill.agents.controllers.pd_joint_pos.PDJointPosController(**kwargs)

Bases: mani_skill.agents.controllers.base_controller.BaseController

Base class for controllers. The controller is an interface for the robot to interact with the environment.

_get_joint_limits()

_initialize_action_space()

before_simulation_step()

Called before each simulation step in one control step.

get_state()

Get the controller state.

Return type:

dict

reset()

Resets the controller to an initial state. This is called upon environment creation and each environment reset

set_action(action)

Set the action to execute. The action can be low-level control signals or high-level abstract commands.

Parameters:

action (torch.Tensor)

set_drive_property()

Set the joint drive property according to the config.

set_drive_targets(targets)

set_state(state)

Parameters:

state (dict)

_start_qpos: torch.Tensor

_target_qpos: torch.Tensor

config: PDJointPosControllerConfig

sets_target_qpos = True

Whether the controller sets the target qpos of the articulation

sets_target_qvel = False

Whether the controller sets the target qvel of the articulation

class mani_skill.agents.controllers.pd_joint_pos.PDJointPosControllerConfig

Bases: mani_skill.agents.controllers.base_controller.ControllerConfig

controller_cls

damping: float | Sequence[float] | numpy.ndarray

drive_mode: Sequence[mani_skill.utils.structs.types.DriveMode] | mani_skill.utils.structs.types.DriveMode = 'force'

force_limit: float | Sequence[float] | numpy.ndarray = 10000000000.0

friction: float | Sequence[float] | numpy.ndarray = 0.0

interpolate: bool = False

lower: None | float | Sequence[float] | numpy.ndarray

The lower bound for the joint position. If none, the lower bound defaults to the lower joint limit.

normalize_action: bool = True

stiffness: float | Sequence[float] | numpy.ndarray

upper: None | float | Sequence[float] | numpy.ndarray

The upper bound for the joint position. If none, the upper bound defaults to the upper joint limit.

use_delta: bool = False

use_target: bool = False

class mani_skill.agents.controllers.pd_joint_pos.PDJointPosMimicController(**kwargs)

Bases: PDJointPosController

Base class for controllers. The controller is an interface for the robot to interact with the environment.

__repr__()

_get_joint_limits()

_initialize_joints()

set_action(action)

Set the action to execute. The action can be low-level control signals or high-level abstract commands.

Parameters:

action (torch.Tensor)

config: PDJointPosMimicControllerConfig

class mani_skill.agents.controllers.pd_joint_pos.PDJointPosMimicControllerConfig

Bases: PDJointPosControllerConfig

PD Joint Position mimic controller configuration. This kind of controller is used to emuluate a mimic joint. For some simulation backends mimic joints are not well simulated and/or are hard to tune and so a mimic controller can be used for better stability as well as alignment with a real robot.

The joint_names field is expected to be a list of all the joints, whether they are the joints to be controlled or the joints that are controlled via mimicing.

mimic is a dictionary that maps the mimic joint name to a dictionary of the format dict(joint: str, multiplier: float, offset: float). joint is the controlling joint name and is required. All given joints in joint_names must be in the mimic dictionary as either a key or a value. The ones that are keys are referred to as the mimic joints and the ones that are values are referred to as the control joints. Mimic joints are the ones directly controlled by the agent/user and control joints are implicitly controlled by following the mimic joints via the following equation:

q_mimic = q_controlling * multiplier + offset

By default, the multiplier is 1.0 and the offset is 0.0.

controller_cls

mimic: dict[str, MimicConfig]

float, offset: float)

Type:

the mimic targets. Maps the actual mimic joint name to a dictionary of the format dict(joint

Type:

str, multiplier