isaaclab_newton.envs.mdp

isaaclab_newton.envs.mdp#

Newton-specific MDP components.

Classes#

The following classes are part of the public isaaclab_newton.envs.mdp API.

NewtonInverseKinematicsAction

Newton inverse-kinematics action term.

NewtonInverseKinematicsActionCfg

Configuration for a Newton inverse-kinematics action term.
class isaaclab_newton.envs.mdp.NewtonInverseKinematicsAction[source]#

Bases: ActionTerm

Newton inverse-kinematics action term.

Solves IK as a single list of objectives on the cloner’s single-env Newton prototype model, then maps the actuated joint coordinates back to the live batched articulation. Each pose objective drives one end-effector body (one is single-body IK, several are multi-body); constraint objectives add no action dimensions. The per-step target computation, seed assembly, solve and gather run entirely in Warp – Torch appears only as the policy action at the boundary, viewed zero-copy into Warp. Fixed-base articulations only.

Methods:

__init__(cfg, env)

Initialize the action term.

__new__(*args, **kwargs)

__init__(cfg: NewtonInverseKinematicsActionCfg, env: ManagerBasedEnv)[source]#

Initialize the action term.

Parameters:

  • cfg – The configuration object.

  • env – The environment instance.

classmethod __new__(*args, **kwargs)#
class isaaclab_newton.envs.mdp.NewtonInverseKinematicsActionCfg[source]#

Bases: ActionTermCfg

Configuration for a Newton inverse-kinematics action term.

The action solves IK as a single list of objectives. Pose objectives (NewtonIKPoseObjectiveCfg) are command-driven and contribute action dimensions – one drives a single-body solve, several drive a multi-body solve. Constraint objectives such as NewtonIKJointLimitObjectiveCfg add residuals but no action dimensions. The action vector is the concatenation of every pose objective’s slice, in list order.

The action currently supports fixed-base articulations only. Each pose objective’s body and the configured joints must resolve both in Isaac Lab and in the registered Newton prototype model for the controlled asset.

Methods:

__new__(*args, **kwargs)

__init__([class_type, asset_name, ...])

classmethod __new__(*args, **kwargs)#
__init__(class_type: type[NewtonInverseKinematicsAction] | str = <factory>, asset_name: str = <factory>, debug_vis: bool = <factory>, clip: dict[str, tuple] | None = <factory>, joint_names: list[str] = <factory>, objectives: list[NewtonIKObjectiveCfg] = <factory>, controller: NewtonIKSolverCfg = <factory>) → None#