Skip to main content

Get Started with Physical AI

Use this guide when an engineering team is preparing its first Physical AI workflow in FactVerse. The goal is to turn digital twin scenes, model assets, geometry, process data, operating constraints, and review evidence into reusable scenario packages.

Setup Flow​

Prerequisites​

RequirementWhy it matters
Scenario ownerDefines the engineering question and review criteria.
Model asset ownerConfirms geometry, version, coordinate system, and source file.
Data ownerPrepares process rules, observations, limits, and operational data.
Runtime ownerConfirms whether the scenario uses FactVerse, SimRunner, Omniverse, Isaac, PhysX, Newton, or another environment.

Inputs​

InputPreparation
Model assetsSource file, version, metadata, component geometry, and twin binding.
Scene contextSpace, equipment, layout, task sequence, safety zones, and coordinate assumptions.
Operational dataProcess timing, production rules, meter values, inspection evidence, and constraints.
Simulation settingsEngine, duration, random seed, changed parameters, and output format.
Review evidenceKPIs, run records, semantic scene JSON, screenshots, notes, and reviewer.

Procedure​

  1. Select one scenario and define the question being reviewed.
  2. Prepare model assets, component geometry, and twin bindings.
  3. Prepare governed data and operating constraints.
  4. Choose the right study type: layout, DES, what-if, building simulation, SimRunner, or external runtime handoff.
  5. Run the study and record assumptions.
  6. Package outputs with evidence and reviewer notes.
  7. Use feedback to revise geometry, data, constraints, or runtime settings.

Expected Output​

The first workflow should produce a reviewed scenario package with stable identity, source data, assumptions, run outputs, and validation notes.

Validation Checklist​

  • Scenario question, owner, and review criteria are explicit.
  • Model asset version and coordinate assumptions are recorded.
  • Source data and constraints are linked to governed records.
  • Simulation settings and runtime are captured.
  • Outputs are stored with reviewer notes.

Failure Handling​

SymptomResponse
Geometry and site coordinates disagreePause study and correct units, origin, orientation, or floor reference.
Scenario lacks process dataKeep output as a visual review until process constraints are prepared.
Runtime choice is unclearUse a FactVerse review workflow until target runtime owner confirms requirements.
Validation evidence is missingKeep the scenario in review status.