Physical System Workspace
Organize a System into components, robot assemblies, typed properties and system-level specifications.
Features / Engineering platform
Define. Simulate.
Understand.
A coherent physical engineering workflow. Product capabilities below distinguish architecture and implemented source from planned desktop validation.
01 / System construction
Model the system,
not just the picture.
3D digital twins
Represent physical geometry, coordinate frames, structures and the relationships between model parts. Native editor maturity is not yet verified.
Actuation and sensing
Define sensors, joint actuation, power, latency and other measurable hardware-related parameters.
Physical constraints
Describe payload, motion limits, thermal envelopes and safety-related preconditions explicitly.
Environment setup
Assemble a scenario's world conditions, objects, physical assumptions and initial state.
System versions
Track intended world revisions, run manifests, observations and exportable evidence.
02 / Experiment lifecycle
Clear inputs.
Inspectable outputs.
Source-level simulation models and evidence concepts exist in the repository's historical simulation stack. That is not equivalent to a verified, publicly installed desktop application or a validated external dynamics engine.
Product-status boundary: The site presents an engineering direction. Native desktop solver execution, reliable model-format import and end-to-end packaged workflows remain to be proven.
- Specify. Capture robot, component, parameter and scenario contracts.
- Pre-check. Evaluate deterministic readiness and parameter constraints, without representing this as physical validation.
- Run. Use only a solver adapter that has actually passed integration tests.
- Trace. Preserve events, state, run metadata and numerical outputs for review.
- Revise. Change a model, rerun and compare reproducible results.
03 / Research horizon
Future research and
integration targets.
MuJoCo, Gazebo, Isaac Sim
Candidate integration directions; no claim that all bridges, model conversions or required GPU features currently work.
Kinematics and contact
Humanoid motion and contact experiments need selected, tested solvers and traceable parameterization.
Sim-to-real studies
Later hardware-facing evaluation and verification. Simulation evidence is not real-world safety certification.