Skip to content

incident-report.md

Latest commit

 

History

History
66 lines (47 loc) · 4.01 KB

incident-report.md

File metadata and controls

66 lines (47 loc) · 4.01 KB
Raw

Incident Report

Formal artifact: Cell Truth Report.

This report summarizes one recorded run: a saved sequence of object positions and timestamps from a bounded workcell.

What happened

  • Metriplane reviewed 8.0 s of recorded activity.
  • It recorded 4 events and grouped them into 1 incident.
  • An event is one observed change, such as an object arriving or a required tool being absent.
  • An incident groups related events that need review because an expected process condition was not met.
  • Incident INC-0001 (warning): amr 2 missing during Second AMR reaches the rendezvous zone. Second AMR reaches the rendezvous zone waited because amr 2 was absent.
  • Suggested follow-up: Add a required-tool staging check. INC-0001 shows a required asset was absent while a process step waited. Add a visible pre-step tool/material check.

Why it was flagged

Process rules describe the expected steps, required objects and locations, and allowed waiting times. Metriplane flags a deviation when the recorded run does not meet one of those rules.

reason severity expected step object supporting event records
missing required asset (missing_required_asset) warning two_amr_rendezvous amr_2 evt_0001, evt_0002

When it happened

The times below are measured from the start of the recorded run.

time what was observed object technical event record
2.0 s amr_2 was not present for Second AMR reaches the rendezvous zone. amr_2 evt_0001 / required_asset_missing
5.0 s Second AMR reaches the rendezvous zone waited 3.0 s for amr_2. amr_2 evt_0002 / step_delayed
6.0 s Synthetic required AMR was present for Second AMR reaches the rendezvous zone. amr_2 evt_0003 / required_asset_present
6.0 s Second AMR reaches the rendezvous zone completed. amr_2 evt_0004 / step_completed

Evidence that was saved

An evidence bundle is a checksummed ZIP that keeps the incident report and supporting records together so another person can verify what was reviewed.

  • Evidence bundle for INC-0001: evidence_bundles/INC-0001.zip
  • Event records: physical_event_log.jsonl
  • Incident records: incidents.jsonl
  • Supporting recorded state: state_segment.jsonl
  • Formal Cell Truth Report: cell_truth_report.html

Repeatable check that was generated

A regression check is a generated test that replays the saved incident and checks that the expected events and incident still appear within declared timing tolerances.

Deterministic replay uses the saved inputs and recorded time sequence instead of live timing, so repeated evaluations can be compared consistently.

  • Repeatable check for INC-0001: regression_tests/INC-0001.yaml

External fixture provenance

  • Fixture: massrobotics_amr_synthetic_incident_v1.
  • Contract: metriplane.external_source_contract.v1 / metriplane.atlas.complete_snapshot.v1.
  • Source: Metriplane synthetic MassRobotics-format engineering source at revision git_commit:6a24ebf1bda47860bda9c5a9bca2c0d94eb37b15.
  • Adapter: org.metriplane.massrobotics_amr_offline_replay version 1.0.0 at commit 6a24ebf1bda47860bda9c5a9bca2c0d94eb37b15.
  • Full conversion provenance: external_source_provenance.json (SHA-256 1f0f1432aa48a3d36e39fc6f1519b3bbda51080756981b62c6f998eddc4f0b61).
  • This identifies the supplied normalized fixture and its conversion; the incident result still comes from the recorded normalized state and the supplied process rules.

Limits of this result

  • This result is derived from recorded normalized planar object state supplied for evaluation, not a fresh interpretation of raw sensor data.
  • It depends on the object identities, state, zone assignments, and process rules supplied for this run.
  • It does not establish correctness, calibration, physical accuracy, or simulator realism of the upstream state.
  • It is not a certified safety or quality decision and does not control machinery.
  • It does not prove root cause; suggested follow-ups require before-and-after validation.