Workcell black box
A replayed workcell incident should become evidence, not another raw log.
Metriplane turns replayed or calibrated workcell state into a physical event ledger, Cell Truth Report, portable evidence bundle, local verification result, and generated regression test.
Built for teams that need incident context they can inspect, replay, verify, and keep as a regression check.
Release evidence Verified release values
Evidence workflow
The product surface is the chain of custody.
Metriplane does not only detect that something happened. It packages a physical incident into a durable artifact that can be inspected, verified, and reused as a regression test.
- 01 replayed workcell state
- 02 physical event
- 03 incident
- 04 evidence bundle
- 05 verification
- 06 generated regression test
Generated artifacts
The output is designed for review, reproduction, and future tests.
Robotics teams already use rosbags, logs, traces, dashboards, and simulation replays. Metriplane formalizes the next layer: incident context that can be bundled, verified, and promoted into a regression check.
What it offers
A bounded evidence layer for robotics incident review.
Turn replayed state into an incident record.
Replay-derived physical events become a dated, inspectable timeline with process deviations and incident IDs.
Package the incident as portable evidence.
INC-0001 carries its manifest, checksums, incident JSON, event timeline, report, replay command, and limitations.
Keep the lesson as a regression test.
The missing-tool incident becomes missing_tool_caused_delay_INC-0001 and passes as a generated physical regression check.
Public demos
Watch the evidence workflow, then inspect the lineage.
The current demo is the SoftwareX artifact path: replayed assembly-cell state becomes an incident, evidence bundle, verification result, and generated regression test. The earlier demos stay visible as project lineage, not as unsupported current-release claims.
current release evidence loop Demo 08
Earlier public demos Lineage from camera tracking to evidence review.
Clean install, operator UI, live fusion, evidence export
The public release path shows setup, operator-facing UI, live fused state, and exported evidence artifacts.
Watch on YouTubeBackpressure and health degradation
The system demonstrates operational pressure, degradation signals, and health behavior under load.
Watch on YouTubeDeterministic replay and run provenance
Replayed runs preserve provenance so results can be compared, inspected, and repeated.
Watch on YouTubeEnd-to-end digital twin with multi-camera fusion
ROS 2 and Omniverse connect the physical scene, fused observations, and a live digital twin.
Watch on YouTubeMulti-camera world-coordinate sensor fusion
Multiple camera feeds are reconciled into shared world coordinates for a coherent cell view.
Watch on YouTubeFrom camera observations to ROS 2
The early pipeline turns observed marker state into ROS 2 messages and live system state.
Watch on YouTubeArUco-driven real-time digital twin in Omniverse
The first demo connects tagged objects to a live Omniverse digital twin.
Watch on YouTubeClaim to evidence
Public claims are mapped to concrete artifacts.
Each proof point should be inspectable through the repository, release archive, evidence package, or camera-free reproduction path.
| Claim | Evidence | How to verify | Boundary |
|---|---|---|---|
| v0.2.0 is archived as a public release | GitHub release and Zenodo DOI 10.5281/zenodo.20736619 | Open the v0.2.0 release and Zenodo record. | SoftwareX acceptance or peer review is not claimed. |
| The release gate passed locally | evidence/paper_v2_0/test_output.txt reports 580 passed. | Run PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/python -m pytest -q. | Captured in one local release environment. |
| The camera-free replay is deterministic | deterministic_replay.txt reports pass=true, 0.0 cm position difference, and 0 event mismatches. | Run RUNS=evidence/paper_v2_0/runs ./tools/mp.sh deterministic-replay datasets/demo/session_001.jsonl. | Checked-in demo session only; no live-camera claim. |
| The assembly-cell replay produces one incident | atlas_assembly_cell_run.txt reports events=6 and incidents=1. | Run metriplane atlas run with the assembly_cell domain pack. | One deterministic assembly-cell domain pack and replay. |
| INC-0001 is explainable as a Cell Truth Report | cell_truth_report.md records a 35.0 s wait for torque_driver_1. | Inspect evidence/paper_v2_0/atlas_run/cell_truth_report.md. | Derived from replayed planar state, not raw-video judgement. |
External review
Feedback is collected as technical review, not social proof.
Public feedback surfaces include GitHub Issue #6, the technical feedback form, an NVIDIA Developer Forum thread, and a DEV.to article. The site does not publish private quotes without permission.
Raw evidence layer
Robotics practitioners already use rosbags, logs, and traces as raw incident evidence.
Structured next layer
Metriplane is positioned as incident context -> bundle -> verification -> regression test.
Permission policy
Named quotes require permission. Otherwise feedback is summarized by theme.
Scope and non-claims
The project is intentionally bounded.
Metriplane is observe-only, local-first, camera/replay-oriented, planar/tagged-asset scoped, and research-software oriented.
Observe-only Planar/tagged-asset scoped Camera-free replay path No robot or machine controlNo safety certificationNo quality-release approvalNo people recognitionNo marker-free tracking claimNo full 3D reconstruction claimNo production-factory validationNo factory-wide deployment readiness
Next step
Inspect the artifact or send technical feedback.
Metriplane improves fastest when reviewers run the camera-free path, challenge the boundaries, and point out missing evidence.