API¶

This page lists every public symbol exported by robometrics.__all__.

Symbol	Module	Description
AgentState	robometrics.schemas	Planar constant-velocity agent state.
BenchmarkProfile	robometrics.benchmarks	Declarative benchmark profile metadata.
CheckpointEntry	robometrics.history	One recorded evaluation checkpoint.
ComparisonResult	robometrics.results	Metric-by-metric comparison between two evaluation results.
DatasetValidationResult	robometrics.validation	Validation report for one CSV or JSON trajectory-style dataset.
EvaluationInputError	robometrics.evaluator	Raised when an evaluation request cannot be constructed.
EvaluationHistory	robometrics.history	Track EvaluationResult objects across training checkpoints.
EvaluationResult	robometrics.results	Collection of metric results from one local evaluation run.
EVALUATION_RESULT_SCHEMA_VERSION	robometrics.results	Stable EvaluationResult JSON schema version string.
Evaluator	robometrics.evaluator	Evaluate registered RoboMetrics metrics against local Python inputs, including Trajectory schema objects.
MetricDefinition	robometrics.registry	Registered metric metadata and execution contract.
MetricComparison	robometrics.results	Comparison of one metric across two evaluation results.
MetricAccumulator	robometrics.accumulator	Stateful per-metric accumulator for streaming evaluation.
MetricRegistry	robometrics.registry	Registry for named metric callables.
MetricResult	robometrics.results	Result for one metric execution.
Trajectory	robometrics.schemas	Serializable trajectory with optional timestamps.
TrajectoryIOError	robometrics.io	Raised when a trajectory file cannot be read or parsed.
UnknownMetricError	robometrics.registry	Raised when a metric name is not registered.
ValidationIssue	robometrics.validation	One dataset validation issue.
__version__	robometrics._version	Package version string.
acceleration	robometrics.comfort	Return approximate acceleration vectors from a position trajectory.
acceleration_magnitude	robometrics.comfort	Return per-step acceleration magnitudes.
acceleration_limits_violated	robometrics.physics	Return a MetricResult describing whether acceleration exceeds max_accel.
action_jerk	robometrics.temporal	Return mean squared second finite difference of an action sequence.
ade	robometrics.trajectory	Return mean pointwise Euclidean distance between predicted and ground-truth paths.
average_displacement_error	robometrics.trajectory	Return mean pointwise Euclidean distance between predicted and ground-truth paths.
behavioral_diversity	robometrics.diversity	Return mean pairwise distance between unique behavior samples.
calibration_error	robometrics.calibration	Return expected calibration error for binary outcomes and confidences.
collision_rate	robometrics.safety	Return fraction of actor-covered ego timesteps that collide with at least one actor.
collision_rate_obb	robometrics.safety	Return fraction of actor-covered ego timesteps with OBB collision.
compounding_error_index	robometrics.temporal	Return normalized non-negative temporal error growth.
contact_richness	robometrics.manipulation	Return fraction of timesteps with meaningful contact force magnitude.
control_smoothness	robometrics.temporal	Return a bounded control smoothness score.
coverage_score	robometrics.coverage	Return occupied grid-bin coverage as a score in [0, 1].
curvature	robometrics.trajectory	Return approximate planar XY curvature assuming uniformly spaced samples.
curvature_profile	robometrics.trajectory	Return approximate planar XY curvature assuming uniformly spaced samples.
curvature_limits_violated	robometrics.physics	Return a MetricResult describing whether curvature exceeds max_curvature.
displacement_at_k	robometrics.driving	Return mean displacement error using only the top-K predicted modes.
dynamic_feasibility	robometrics.physics	Return a Newtonian dynamic feasibility score.
dynamic_feasibility_score	robometrics.physics	Return a conservative 0..1 feasibility score against dynamic limits.
end_effector_tracking_error	robometrics.manipulation	Return mean Euclidean distance between end-effector and target path.
failure_severity	robometrics.safety	Return aggregated numeric severity for failure events.
fde	robometrics.trajectory	Return final-step Euclidean distance between predicted and ground-truth paths.
final_displacement_error	robometrics.trajectory	Return final-step Euclidean distance between predicted and ground-truth paths.
force_limit_compliance	robometrics.manipulation	Return fraction of timesteps where force magnitude is within limit.
get_profile	robometrics.benchmarks	Return one benchmark profile by name.
goal_reaching_accuracy	robometrics.task	Return fraction of goals reached within Euclidean tolerance.
grasp_success_rate	robometrics.manipulation	Return fraction of grasp attempts that succeeded.
hausdorff_distance	robometrics.trajectory	Return symmetric Hausdorff distance using all coordinate dimensions.
intervention_free_time	robometrics.safety	Return duration of intervention-free segments.
jerk	robometrics.comfort	Return approximate jerk vectors from a position trajectory.
jerk_cost	robometrics.comfort	Return mean squared jerk magnitude.
jerk_magnitude	robometrics.comfort	Return per-step jerk magnitudes.
jerk_limits_violated	robometrics.physics	Return a MetricResult describing whether jerk exceeds max_jerk.
joint_limit_violation_rate	robometrics.manipulation	Return fraction of configs where any joint exceeds its limits.
kinematic_feasibility	robometrics.physics	Return a trajectory kinematic feasibility score.
lane_departure_rate	robometrics.safety	Return fraction of ego points outside a polygonal lane boundary.
lateral_error	robometrics.trajectory	Return mean absolute planar XY lateral deviation from a reference trajectory.
list_profiles	robometrics.benchmarks	Return all built-in benchmark profiles.
load_pack	robometrics.registry	Import a metric pack module and register its metrics.
load_trajectory	robometrics.io	Load a trajectory from .npy, .npz, .csv, or .json.
long_horizon_drift	robometrics.temporal	Return later-weighted mean rollout error.
longitudinal_error	robometrics.trajectory	Return mean absolute planar XY longitudinal deviation along a reference trajectory.
max_acceleration	robometrics.comfort	Return maximum acceleration magnitude.
max_deceleration	robometrics.comfort	Return maximum longitudinal deceleration magnitude.
mean_acceleration	robometrics.comfort	Return mean acceleration magnitude.
mean_curvature	robometrics.trajectory	Return mean planar XY curvature.
min_ade	robometrics.prediction	Return the minimum average displacement error over predicted modes.
min_distance_to_actors	robometrics.safety	Return minimum time-aligned Euclidean distance from ego to any actor.
min_fde	robometrics.prediction	Return the minimum final displacement error over predicted modes.
miss_rate	robometrics.prediction	Return a per-sample miss indicator as 0.0 or 1.0.
near_miss_rate	robometrics.safety	Return fraction of events that are near misses without collision.
offroad_rate	robometrics.driving	Return fraction of ego positions outside all drivable-area polygons.
path_length	robometrics.trajectory	Return total Euclidean path length.
physics_violation_rate	robometrics.physics	Return fraction of timesteps with any physics violation.
prediction_nll	robometrics.driving	Return mean negative log-likelihood for a Gaussian mixture prediction.
recovery_success_rate	robometrics.safety	Return successful recoveries divided by recovery opportunities.
registry	robometrics.registry	Registry for named metric callables.
rms_acceleration	robometrics.comfort	Return root-mean-square acceleration magnitude.
run_profile	robometrics.benchmarks	Run a benchmark profile against prediction and ground-truth arrays.
smoothness_score	robometrics.comfort	Return a scale-normalized third-difference smoothness score.
soft_ttc	robometrics.driving	Return minimum constant-velocity TTC computed at each rollout timestep.
speed_profile	robometrics.physics	Return speed magnitude at each trajectory point.
task_success_rate	robometrics.task	Return fraction of task attempts that succeeded.
temporal_drift	robometrics.temporal	Return mean slope of L2 error over time.
time_to_collision	robometrics.safety	Return constant-velocity time to collision for two disc agents.
topk_trajectory_error	robometrics.prediction	Return best ADE among the first k predictions, assumed confidence-ranked.
trajectory_diversity	robometrics.diversity	Return mean pairwise ADE between predicted trajectory modes.
validate_dataset	robometrics.validation	Validate a trajectory-style CSV or JSON file.
workspace_coverage	robometrics.coverage	Return the number of occupied grid cells visited by sampled positions.