How would you validate a replayed workcell incident as an evidence bundle and regression test?

Omniverse Isaac Sim
, , ,
1

I am not reporting an Isaac Sim bug. I am looking for technical feedback from people working with robotics simulation, Isaac Sim, Omniverse, digital twins, and workcell validation.

I released MetriPlane v0.2.0, an open-source, observe-only physical-observability tool for bounded workcells.

3-minute demo:

Repository:

Short 2–4 minute feedback form:

Core workflow

The v0.2.0 demo is camera-free and replay-based.

It shows a missing torque-driver event becoming:

replayed / calibrated workcell state
→ physical event log
→ Cell Truth Report
→ incident evidence bundle
→ local bundle verification
→ generated regression test
→ regression PASS

The idea is that a physical incident should not remain only a log line. It should become a reviewable, replayable, and testable artifact.

My question

For people working with Isaac Sim, Omniverse, robotics simulation, or digital-twin workflows:

Does this “incident → evidence bundle → regression test” workflow make sense as a simulation / workcell-validation pattern?

I am especially interested in:

  1. Would this be useful around simulation validation, digital twins, physical-AI testing, or workcell review?

  2. What would be the most important missing validation step?

  3. What should be added before this is useful beyond a deterministic demo fixture?

  4. Is the observe-only boundary clear enough?

  5. Would an Isaac / Omniverse integration be valuable as an export/replay surface, or is the current framing wrong?

Current v0.2.0 demo evidence

The current demo shows:

6 physical events
1 incident
35.0 second missing-tool delay
INC-0001 evidence bundle
bundle verify: pass=true
generated regression test: PASS

Scope and boundaries

MetriPlane v0.2.0 is intentionally bounded.

It is:

  • observe-only

  • local-first

  • camera/replay-oriented

  • planar/tagged-asset scoped

  • research-software oriented

It does not claim:

  • robot or machine control

  • safety certification

  • quality-release approval

  • marker-free tracking

  • full 3D reconstruction

  • factory-wide production deployment

  • Isaac Sim runtime validation in the current artifact

The current Isaac / Omniverse relevance is mostly architectural: I am trying to understand whether this evidence-bundle / regression-test loop would be useful around simulation validation and digital-twin review workflows.

Critical feedback is preferred.

1 Like