1 rollouts.inspection.constraint_audit

rollouts.inspection.constraint_audit

Descriptive motion-bound opportunity on immutable factual candidate shells.

A motion-passing row is an upper bound on opportunity, never a safe rescued view: bounds, clearance, free-space, path and labels may be unavailable. States and candidates are dependent; summaries average state fractions within physical scenes, then weight scenes equally within each collection and split.

1.1 Functions

Name	Description
reduce_constraint_rows	Reduce uniquely identified factual rows, preserving unavailable diagnostics.

1.1.1 reduce_constraint_rows

rollouts.inspection.constraint_audit.reduce_constraint_rows(rows)

Reduce uniquely identified factual rows, preserving unavailable diagnostics.

Required inputs include cohort/frozen split/scene, full identity, family, observed masks, motion measurements, and optional active yaw/backward caps. Threshold flags use the producer’s strict greater-than rejection predicate. Missing step/height or active yaw/backward diagnostics never count as passing. The grid requires enforced baseline step/height bounds of 1 m / 0.25 m.

# rollouts.inspection.constraint_audit { #aria_nbv.rollouts.inspection.constraint_audit }

`rollouts.inspection.constraint_audit`

Descriptive motion-bound opportunity on immutable factual candidate shells.

A motion-passing row is an upper bound on opportunity, never a safe rescued
view: bounds, clearance, free-space, path and labels may be unavailable. States
and candidates are dependent; summaries average state fractions within physical
scenes, then weight scenes equally within each collection and split.

## Functions

| Name | Description |
| --- | --- |
| [reduce_constraint_rows](#aria_nbv.rollouts.inspection.constraint_audit.reduce_constraint_rows) | Reduce uniquely identified factual rows, preserving unavailable diagnostics. |

### reduce_constraint_rows { #aria_nbv.rollouts.inspection.constraint_audit.reduce_constraint_rows }

```python
rollouts.inspection.constraint_audit.reduce_constraint_rows(rows)
```

Reduce uniquely identified factual rows, preserving unavailable diagnostics.

Required inputs include cohort/frozen split/scene, full identity, family,
observed masks, motion measurements, and optional active yaw/backward caps.
Threshold flags use the producer's strict greater-than rejection predicate.
Missing step/height or active yaw/backward diagnostics never count as passing.
The grid requires enforced baseline step/height bounds of 1 m / 0.25 m.