System / Component / Robot
Typed physical entities and assembly relationships.
Technology / Engineering platform
Evidence is
the interface.
The architecture aims to connect typed physical models, scenario definitions, simulation computation and auditable run outputs. Roadmap engine names are not proof of working adapters.
01 / Model vocabulary
Sensor / Actuator / Constraint
Measurable signals, operational limits and model assumptions.
Scenario / State / WorldRevision
Experiment configuration and explicit system states.
SimulationRun / Manifest / Trace
Input provenance, events and reproducibility metadata.
Result / Export
Structured outputs with clearly documented fidelity.
Engine adapter
A real computational backend, only after verified integration.
02 / Solver truth
Target engines are
not shipped engines.
Solver compatibility must be demonstrated with actual installs, model translation checks and numerical evidence.
| Technology | Intended role | Public status |
|---|---|---|
| MuJoCo | Candidate articulated-body dynamics adapter | Integration not verified |
| Gazebo | Candidate robotics simulation ecosystem | Integration not verified |
| Isaac Sim | Candidate high-fidelity simulation tooling | Integration not verified |
| Drake / Bullet | Possible research/dynamics interfaces | Integration not verified |
| Readiness checks | Deterministic constraints and budget evaluation in historical source | Not equivalent to physically validated solver dynamics |
Compatibility disclaimer: URDF, SDF, USD and MJCF schema/preview support does not imply imports will execute correctly in every engine. Simulation is never an automatic guarantee of physical-world safety.
03 / Research horizon
Embodied learning and sim-to-real, later.
Policy training, VLA workflows, connected hardware validation and large-scale simulation infrastructure remain research and roadmap objectives. Their readiness cannot be inferred from architectural contracts.
Review the engineering workflow