Skip to content

NATIVE_ENGINE.md

Latest commit

 

History

History
56 lines (37 loc) · 2.01 KB

NATIVE_ENGINE.md

File metadata and controls

56 lines (37 loc) · 2.01 KB

NativeEngine

NativeEngine is the C++ runtime for physics and environment simulation.

Responsibilities

  • Rigid body simulation and constraints.
  • Environment parameters (wind, ambient temperature).
  • Sensor signal generation as needed by the simulation.

Integration

  • NativeEngine is driven from Unity via RobotWin/Assets/Scripts/Core/NativeBridge.cs.
  • Physics stepping is done via NativeBridge.Physics_Step(dt) (see RobotWin/Assets/Scripts/Game/NativePhysicsWorld.cs).
  • Circuit/IO stepping uses NativeBridge.Native_Step(dt) and is kept separate from physics stepping.

Configuration surface

Runtime config is passed via an interop struct (solver iterations, air density, wind, etc.) and is authored on the Unity side (see NativeBridge.PhysicsConfig).

Exported APIs (Unity P/Invoke)

The entry points are declared in RobotWin/Assets/Scripts/Core/NativeBridge.cs.

Circuit API (electrical graph):

  • Native_CreateContext(), Native_DestroyContext()
  • Native_AddNode()
  • Native_AddComponent(type, paramCount, parameters)
  • Native_Connect(compId, pinIndex, nodeId)
  • Native_Step(dt)
  • Native_GetVoltage(nodeId)

Physics API:

  • World lifecycle: Physics_CreateWorld(), Physics_DestroyWorld(), Physics_SetConfig(ref config)
  • Bodies: Physics_AddBody(ref body), Physics_GetBody(id, out body), Physics_SetBody(id, ref body)
  • Stepping: Physics_Step(dt)
  • Vehicles: Physics_AddVehicle(...), Physics_SetWheelInput(...), Physics_SetVehicleAero(...), Physics_SetVehicleTireModel(...)
  • Forces/constraints: Physics_ApplyForce(...), Physics_ApplyForceAtPoint(...), Physics_ApplyTorque(...), Physics_AddDistanceConstraint(...)
  • Queries: Physics_Raycast(...)

Build

  • Use python tools/rt_tool.py build-native.
  • Outputs land in builds/native/.

Interop

  • Exposes a C ABI used by CoreSim and Unity.
  • Inputs and outputs are fixed-size structs to avoid per-step allocations.

Determinism

  • Fixed step sizes in deterministic mode.
  • Avoid nondeterministic sources in hot paths.