embodichain.lab.sim.atomic_actions.tracking

Typed tracking keeps command projection, versioned feedback routing, and metric evaluation separate from command transport. Policies can use these contracts for in-flight recovery checks and terminal acceptance without mixing joint, base-pose, or whole-body error units.

Typed, transport-neutral tracking contracts for atomic-action execution.

Channels and endpoint routes

JOINT_POSITION_CHANNEL	str(object='') -> str str(bytes_or_buffer[, encoding[, errors]]) -> str
BASE_POSE_CHANNEL	str(object='') -> str str(bytes_or_buffer[, encoding[, errors]]) -> str
WHOLE_BODY_POSE_CHANNEL	str(object='') -> str str(bytes_or_buffer[, encoding[, errors]]) -> str
TrackingChannelId	Open string identifier for one typed endpoint-feedback channel.
TrackingFeedbackAddress	Immutable address understood by one tracking-feedback provider.
EndpointTrackingFeedbackAddress	Feedback address for one runtime endpoint and open tracking channel.
TrackingFeedbackSourceRef	Versioned provider route plus one immutable feedback address.
TrackingProjectorRef	Exact version of a command-to-tracking-state projector.
EndpointTrackingChannelBinding	Resolved source and projector for one endpoint tracking channel.

Typed states and metrics

TrackingState	Immutable-by-ownership typed desired or observed tracking state.
JointPositionTrackingState	Batched joint positions with shape (B, D).
PoseTrackingState	Batched homogeneous poses with shape (B, 4, 4).
WholeBodyPoseTrackingState	Batched base poses and joint positions for whole-body tracking.
TrackingMetricCfg	Immutable tolerance configuration dispatched by exact metric ID/revision.
JointPositionTrackingMetric	Maximum absolute joint-error tolerance.
PoseTrackingMetric	Independent translation and rotation tolerances for base pose.
WholeBodyPoseTrackingMetric	Independent base-pose and joint-position tolerances.

Tracking policies and command-aligned targets

InFlightTrackingPolicy	Feedback checks used while a command sequence is still in flight.
FeedbackTerminalAcceptance	Terminal acceptance proven by typed endpoint feedback.
TimedTerminalAcceptance	Explicit terminal acceptance without endpoint feedback.
TerminalAcceptance	Represent a PEP 604 union type
TrackingPolicy	Independent in-flight recovery signal and terminal acceptance contract.
TrackingSetpoint	One endpoint-local desired state and its typed feedback route.
TrackingFrame	Desired endpoint states associated with one command frame.
TimedTrackingSequence	Tracking frames aligned by index with an authoritative command sequence.

Runtime extension points and built-ins

TrackingFeedbackBatch	One synchronized typed observation from an exact feedback source.
TrackingEvaluation	Per-row metric result with unit-preserving component errors.
TrackingFeedbackProvider	Versioned live port that reads one exact tracking source.
TrackingCommandProjector	Versioned pure projector from an endpoint command to desired state.
TrackingMetricEvaluator	Versioned evaluator for one exact metric configuration type.
TrackingFeedbackProviderRegistry	Immutable exact-version feedback-provider registry.
TrackingProjectorRegistry	Immutable exact-version command-projector registry.
TrackingEvaluatorRegistry	Immutable exact-version metric-evaluator registry.
PlanningContextTrackingFeedbackProvider	Built-in provider backed by PlanningContext.robot.
JointPositionTrackingProjector	Built-in projector for joint-position endpoint commands.
JointPositionTrackingEvaluator	Evaluator for JointPositionTrackingMetric.
PoseTrackingEvaluator	Evaluator for PoseTrackingMetric.
WholeBodyPoseTrackingEvaluator	Evaluator for WholeBodyPoseTrackingMetric.
TrackingRuntime	Runtime facade for projecting commands and evaluating typed feedback.

Classes:

EndpointTrackingChannelBinding	Resolved source and projector for one endpoint tracking channel.
EndpointTrackingFeedbackAddress	Feedback address for one runtime endpoint and open tracking channel.
FeedbackTerminalAcceptance	Terminal acceptance proven by typed endpoint feedback.
InFlightTrackingPolicy	Feedback checks used while a command sequence is still in flight.
JointPositionTrackingEvaluator	Evaluator for JointPositionTrackingMetric.
JointPositionTrackingMetric	Maximum absolute joint-error tolerance.
JointPositionTrackingProjector	Built-in projector for joint-position endpoint commands.
JointPositionTrackingState	Batched joint positions with shape (B, D).
PlanningContextTrackingFeedbackProvider	Built-in provider backed by PlanningContext.robot.
PoseTrackingEvaluator	Evaluator for PoseTrackingMetric.
PoseTrackingMetric	Independent translation and rotation tolerances for base pose.
PoseTrackingState	Batched homogeneous poses with shape (B, 4, 4).
TimedTerminalAcceptance	Explicit terminal acceptance without endpoint feedback.
TimedTrackingSequence	Tracking frames aligned by index with an authoritative command sequence.
TrackingChannelId	Open string identifier for one typed endpoint-feedback channel.
TrackingCommandProjector	Versioned pure projector from an endpoint command to desired state.
TrackingEvaluation	Per-row metric result with unit-preserving component errors.
TrackingEvaluatorRegistry	Immutable exact-version metric-evaluator registry.
TrackingFeedbackAddress	Immutable address understood by one tracking-feedback provider.
TrackingFeedbackBatch	One synchronized typed observation from an exact feedback source.
TrackingFeedbackProvider	Versioned live port that reads one exact tracking source.
TrackingFeedbackProviderRegistry	Immutable exact-version feedback-provider registry.
TrackingFeedbackSourceRef	Versioned provider route plus one immutable feedback address.
TrackingFrame	Desired endpoint states associated with one command frame.
TrackingMetricCfg	Immutable tolerance configuration dispatched by exact metric ID/revision.
TrackingMetricEvaluator	Versioned evaluator for one exact metric configuration type.
TrackingPolicy	Independent in-flight recovery signal and terminal acceptance contract.
TrackingProjectorRef	Exact version of a command-to-tracking-state projector.
TrackingProjectorRegistry	Immutable exact-version command-projector registry.
TrackingRuntime	Runtime facade for projecting commands and evaluating typed feedback.
TrackingSetpoint	One endpoint-local desired state and its typed feedback route.
TrackingState	Immutable-by-ownership typed desired or observed tracking state.
WholeBodyPoseTrackingEvaluator	Evaluator for WholeBodyPoseTrackingMetric.
WholeBodyPoseTrackingMetric	Independent base-pose and joint-position tolerances.
WholeBodyPoseTrackingState	Batched base poses and joint positions for whole-body tracking.

class embodichain.lab.sim.atomic_actions.tracking.EndpointTrackingChannelBinding

Resolved source and projector for one endpoint tracking channel.

Methods:

__init__(channel_id, source, projector)
snapshot()	Return an independently owned channel binding.

Attributes:

route_fingerprint

Return the exact channel, source, and projector route identity.

__init__(channel_id, source, projector)

property route_fingerprint: tuple[str, Hashable, str, str]

Return the exact channel, source, and projector route identity.

snapshot()

Return an independently owned channel binding.

Return type:

EndpointTrackingChannelBinding

class embodichain.lab.sim.atomic_actions.tracking.EndpointTrackingFeedbackAddress

Feedback address for one runtime endpoint and open tracking channel.

Methods:

__init__(target, channel_id)

Attributes:

address_fingerprint

Return the endpoint- and channel-scoped address identity.

__init__(target, channel_id)

property address_fingerprint: Hashable

Return the endpoint- and channel-scoped address identity.

class embodichain.lab.sim.atomic_actions.tracking.FeedbackTerminalAcceptance

Terminal acceptance proven by typed endpoint feedback.

Methods:

__init__(metrics[, settle_timeout, ...])

__init__(metrics, settle_timeout=0.0, consecutive_acceptances=1)

class embodichain.lab.sim.atomic_actions.tracking.InFlightTrackingPolicy

Feedback checks used while a command sequence is still in flight.

Methods:

__init__(metrics[, consecutive_violations, ...])

__init__(metrics, consecutive_violations=1, grace_period=0.0)

class embodichain.lab.sim.atomic_actions.tracking.JointPositionTrackingEvaluator

Evaluator for JointPositionTrackingMetric.

Classes:

metric_type

alias of JointPositionTrackingMetric

metric_type

alias of JointPositionTrackingMetric

class embodichain.lab.sim.atomic_actions.tracking.JointPositionTrackingMetric

Maximum absolute joint-error tolerance.

Methods:

__init__([tolerance])

__init__(tolerance=0.05)

class embodichain.lab.sim.atomic_actions.tracking.JointPositionTrackingProjector

Built-in projector for joint-position endpoint commands.

class embodichain.lab.sim.atomic_actions.tracking.JointPositionTrackingState

Batched joint positions with shape (B, D).

Methods:

__init__(positions)
snapshot()	Return an independently owned state snapshot.

Attributes:

batch_size	Return the represented environment count.
device	Return the tensor device.

__init__(positions)

property batch_size: int

Return the represented environment count.

property device: device

Return the tensor device.

snapshot()

Return an independently owned state snapshot.

Return type:

JointPositionTrackingState

class embodichain.lab.sim.atomic_actions.tracking.PlanningContextTrackingFeedbackProvider

Built-in provider backed by PlanningContext.robot.

class embodichain.lab.sim.atomic_actions.tracking.PoseTrackingEvaluator

Evaluator for PoseTrackingMetric.

Classes:

metric_type

alias of PoseTrackingMetric

metric_type

alias of PoseTrackingMetric

class embodichain.lab.sim.atomic_actions.tracking.PoseTrackingMetric

Independent translation and rotation tolerances for base pose.

Methods:

__init__([translation_tolerance, ...])

__init__(translation_tolerance=0.02, rotation_tolerance=0.05)

class embodichain.lab.sim.atomic_actions.tracking.PoseTrackingState

Batched homogeneous poses with shape (B, 4, 4).

Methods:

__init__(poses)
snapshot()	Return an independently owned state snapshot.

Attributes:

batch_size	Return the represented environment count.
device	Return the tensor device.

__init__(poses)

property batch_size: int

Return the represented environment count.

property device: device

Return the tensor device.

snapshot()

Return an independently owned state snapshot.

Return type:

PoseTrackingState

class embodichain.lab.sim.atomic_actions.tracking.TimedTerminalAcceptance

Explicit terminal acceptance without endpoint feedback.

Methods:

__init__([settle_duration])

__init__(settle_duration=0.0)

class embodichain.lab.sim.atomic_actions.tracking.TimedTrackingSequence

Tracking frames aligned by index with an authoritative command sequence.

Methods:

__init__(env_ids, frames)

Attributes:

batch_size	Return the represented environment count.
device	Return the sequence tensor device.
frame_count	Return the number of command-aligned tracking frames.

__init__(env_ids, frames)

property batch_size: int

Return the represented environment count.

property device: device

Return the sequence tensor device.

property frame_count: int

Return the number of command-aligned tracking frames.

embodichain.lab.sim.atomic_actions.tracking.TrackingChannelId

Open string identifier for one typed endpoint-feedback channel.

class embodichain.lab.sim.atomic_actions.tracking.TrackingCommandProjector

Versioned pure projector from an endpoint command to desired state.

Methods:

__init__(*args, **kwargs)
project(command, binding)	Project one command into the binding's desired tracking channel.

__init__(*args, **kwargs)

project(command, binding)

Project one command into the binding’s desired tracking channel.

Return type:

TrackingState

class embodichain.lab.sim.atomic_actions.tracking.TrackingEvaluation

Per-row metric result with unit-preserving component errors.

Methods:

__init__(channel_id, accepted_mask, ...[, ...])

__init__(channel_id, accepted_mask, valid_mask, normalized_error, component_errors=<factory>)

class embodichain.lab.sim.atomic_actions.tracking.TrackingEvaluatorRegistry

Immutable exact-version metric-evaluator registry.

Methods:

__init__([evaluators])

__init__(evaluators=())

class embodichain.lab.sim.atomic_actions.tracking.TrackingFeedbackAddress

Immutable address understood by one tracking-feedback provider.

Attributes:

address_fingerprint

Return a stable, hashable address identity.

Methods:

snapshot()

Return an independently owned address snapshot.

abstract property address_fingerprint: Hashable

Return a stable, hashable address identity.

snapshot()

Return an independently owned address snapshot.

Return type:

TrackingFeedbackAddress

class embodichain.lab.sim.atomic_actions.tracking.TrackingFeedbackBatch

One synchronized typed observation from an exact feedback source.

Methods:

__init__(source, state, valid_mask, timestamp)

__init__(source, state, valid_mask, timestamp)

class embodichain.lab.sim.atomic_actions.tracking.TrackingFeedbackProvider

Versioned live port that reads one exact tracking source.

Methods:

__init__(*args, **kwargs)
observe(source, context)	Read one synchronized typed feedback batch.

__init__(*args, **kwargs)

observe(source, context)

Read one synchronized typed feedback batch.

Return type:

TrackingFeedbackBatch

class embodichain.lab.sim.atomic_actions.tracking.TrackingFeedbackProviderRegistry

Immutable exact-version feedback-provider registry.

Methods:

__init__([providers])

__init__(providers=())

class embodichain.lab.sim.atomic_actions.tracking.TrackingFeedbackSourceRef

Versioned provider route plus one immutable feedback address.

Methods:

__init__(provider_id, revision, address)
snapshot()	Return an independently owned source reference.

Attributes:

source_fingerprint

Return the exact versioned source identity.

__init__(provider_id, revision, address)

snapshot()

Return an independently owned source reference.

Return type:

TrackingFeedbackSourceRef

property source_fingerprint: Hashable

Return the exact versioned source identity.

class embodichain.lab.sim.atomic_actions.tracking.TrackingFrame

Desired endpoint states associated with one command frame.

Methods:

__init__([setpoints])

__init__(setpoints=())

class embodichain.lab.sim.atomic_actions.tracking.TrackingMetricCfg

Immutable tolerance configuration dispatched by exact metric ID/revision.

Methods:

snapshot()

Return an independently owned metric configuration.

snapshot()

Return an independently owned metric configuration.

Return type:

TrackingMetricCfg

class embodichain.lab.sim.atomic_actions.tracking.TrackingMetricEvaluator

Versioned evaluator for one exact metric configuration type.

Methods:

__init__(*args, **kwargs)
evaluate(desired, observed, valid_mask, metric)	Evaluate a desired and observed batch row by row.

__init__(*args, **kwargs)

evaluate(desired, observed, valid_mask, metric)

Evaluate a desired and observed batch row by row.

Return type:

TrackingEvaluation

class embodichain.lab.sim.atomic_actions.tracking.TrackingPolicy

Independent in-flight recovery signal and terminal acceptance contract.

Methods:

__init__(in_flight, terminal)
joint_position(*[, in_flight_max_abs_error, ...])	Create the built-in joint-position tracking and acceptance contract.
timed(*[, settle_duration])	Create an explicit time-only terminal contract with no tracking.

__init__(in_flight, terminal)

classmethod joint_position(*, in_flight_max_abs_error=0.05, terminal_max_abs_error=0.05, terminal_settle_timeout=0.5, consecutive_violations=1, consecutive_acceptances=1, grace_period=0.0)

Create the built-in joint-position tracking and acceptance contract.

Return type:

TrackingPolicy

classmethod timed(*, settle_duration=0.0)

Create an explicit time-only terminal contract with no tracking.

Return type:

TrackingPolicy

class embodichain.lab.sim.atomic_actions.tracking.TrackingProjectorRef

Exact version of a command-to-tracking-state projector.

Methods:

__init__(projector_id, revision)
snapshot()	Return an independently owned projector route.

__init__(projector_id, revision)

snapshot()

Return an independently owned projector route.

Return type:

TrackingProjectorRef

class embodichain.lab.sim.atomic_actions.tracking.TrackingProjectorRegistry

Immutable exact-version command-projector registry.

Methods:

__init__([projectors])

__init__(projectors=())

class embodichain.lab.sim.atomic_actions.tracking.TrackingRuntime

Runtime facade for projecting commands and evaluating typed feedback.

Methods:

__init__(providers, projectors, evaluators)
evaluate(setpoint, feedback, metric)	Evaluate one observed setpoint with an exact metric implementation.
evaluate_frame(frame, metrics, context)	Observe and evaluate every setpoint required by one frame.
observe(setpoint, context)	Read the exact feedback source for one setpoint.
project(command, binding)	Project one command through the exact binding-owned projector.
with_builtins()	Create a runtime with context feedback and built-in typed metrics.

Attributes:

evaluators	Return the immutable exact-version evaluator registry.
projectors	Return the immutable exact-version projector registry.
providers	Return the immutable exact-version provider registry.

__init__(providers, projectors, evaluators)

evaluate(setpoint, feedback, metric)

Evaluate one observed setpoint with an exact metric implementation.

Return type:

TrackingEvaluation

evaluate_frame(frame, metrics, context)

Observe and evaluate every setpoint required by one frame.

Return type:

Mapping[tuple[str, str, str], TrackingEvaluation]

property evaluators: TrackingEvaluatorRegistry

Return the immutable exact-version evaluator registry.

observe(setpoint, context)

Read the exact feedback source for one setpoint.

Return type:

TrackingFeedbackBatch

project(command, binding)

Project one command through the exact binding-owned projector.

Return type:

TrackingState

property projectors: TrackingProjectorRegistry

Return the immutable exact-version projector registry.

property providers: TrackingFeedbackProviderRegistry

Return the immutable exact-version provider registry.

classmethod with_builtins()

Create a runtime with context feedback and built-in typed metrics.

Return type:

TrackingRuntime

class embodichain.lab.sim.atomic_actions.tracking.TrackingSetpoint

One endpoint-local desired state and its typed feedback route.

Methods:

__init__(endpoint_key, binding, desired)

__init__(endpoint_key, binding, desired)

class embodichain.lab.sim.atomic_actions.tracking.TrackingState

Immutable-by-ownership typed desired or observed tracking state.

Attributes:

batch_size	Return the represented environment count.
device	Return the tensor device.

Methods:

snapshot()

Return an independently owned state snapshot.

abstract property batch_size: int

Return the represented environment count.

abstract property device: device

Return the tensor device.

abstract snapshot()

Return an independently owned state snapshot.

Return type:

TrackingState

class embodichain.lab.sim.atomic_actions.tracking.WholeBodyPoseTrackingEvaluator

Evaluator for WholeBodyPoseTrackingMetric.

Classes:

metric_type

alias of WholeBodyPoseTrackingMetric

metric_type

alias of WholeBodyPoseTrackingMetric

class embodichain.lab.sim.atomic_actions.tracking.WholeBodyPoseTrackingMetric

Independent base-pose and joint-position tolerances.

Methods:

__init__([translation_tolerance, ...])

__init__(translation_tolerance=0.02, rotation_tolerance=0.05, joint_position_tolerance=0.05)

class embodichain.lab.sim.atomic_actions.tracking.WholeBodyPoseTrackingState

Batched base poses and joint positions for whole-body tracking.

Methods:

__init__(root_poses, joint_positions)
snapshot()	Return an independently owned state snapshot.

Attributes:

batch_size	Return the represented environment count.
device	Return the tensor device.

__init__(root_poses, joint_positions)

property batch_size: int

Return the represented environment count.

property device: device

Return the tensor device.

snapshot()

Return an independently owned state snapshot.

Return type:

WholeBodyPoseTrackingState