Robots & Learning (MDP)

APIs for authoring robot agents and reinforcement-learning tasks: the MDP component model (observations, rewards, terminations, randomization), robot controllers, and agent management.

Type	Kind	Summary
ActionGroupSlotInfo	class	Describes a resolved action group for multi-policy control.
AgentBatch	class
AgentResetData	struct
BodyData	class
CalibrationMapping	struct
CommandManager	class	Holds a RobotAgent's commands (e.g.
ComponentDescriptor	struct	Describes one available MDP component for the capability manifest.
ContractNegotiator	class	Validates a TaskContract, resolves optional terms, configures the engine's observation/reward/termination pipeline, and returns a NegotiatedTaskSession that subsequent Step/Reset calls reference.
ContractValidator	class	Validates a TaskContract against the MdpComponentRegistry.
ExternalAgentRegistry	static class	Thread-safe bridge between the engine-driven Learn pipeline and the gRPC services.
HzMath	static class
HzObservations	static class
IAuxiliaryDataProvider	interface	Auxiliary data provider.
ICommand	interface
IMdpObservation	interface	Engine-side observation function.
IMdpReward	interface	Engine-side reward signal.
IMdpTermination	interface	Engine-side termination condition.
IRandomizationHandler	interface	Domain randomization handler.
JointConfig	struct
ManifestSnapshot	class	Snapshot of all MDP component descriptors for a given robot.
MdpComponentRegistry	static class	Central registration table for engine-side MDP components.
MdpContext	class	Immutable context passed to MDP component Compute/Evaluate methods.
MdpObservationAttribute	class	Marks a static method as an MDP observation component.
MdpRewardAttribute	class	Marks a static method as an MDP reward component.
MdpScriptDiscovery	static class	Discovers user-defined MDP components from loaded C# assemblies.
MdpTerminationAttribute	class	Marks a static method as an MDP termination component.
NegotiatedSession	class	Runtime state for a negotiated task session.
NegotiationOutcome	class	Result of a negotiation attempt.
ObservationComponents	static class	Built-in observation components that extract robot state from MuJoCo.
ObservationSlotInfo	class	Describes one slot in the negotiated observation layout.
ParamDescriptor	struct	Describes an accepted parameter for an MDP component.
RandomizationDescriptor	struct	Extended descriptor for randomization components with range defaults.
RandomizationHandlers	static class	Built-in domain randomization handlers.
RefPoseCalibrator	class
RefPoseExtractor	class
RewardComponents	static class	Built-in reward signal components computed from MuJoCo state.
RobotAgent	abstract class	Base class for a robot in the Learn pipeline.
RobotController	struct
RobotEnv	class	Internal per-policy batching layer owned by RobotManager.
RobotManager	class	Owns the per-tick Learn pipeline for a set of RobotAgents.
StateConfig	struct
TerminationComponents	static class	Built-in termination conditions evaluated from MuJoCo state.
ValidationMessage	class	Individual validation message with actionable suggestion.
ValidationResult	class	Result of contract validation.
VelocityCommand	class