Skip to content

Capability Model

robot_behavior expresses capabilities through composable traits. A type does not need to implement everything; it exposes only the behavior it actually supports.

Core Layers

Robot
├── Joints<N>
├── EndPoint
├── MoveTo<S>
├── MoveTraj<S>
├── ControlWith<S>
└── Arm<N> = Robot + Joints<N> + EndPoint + MoveTo<JointSpace<N>> + MoveTo<FlangeSpace>

Arm<N> is the aggregate arm trait, but it is not the only entry point. A simulation-only type may implement RobotDescription; a driver supporting one control channel may implement just that ControlWith<S>.

Data and State

  • Robot::State is the driver's native state and may mirror a vendor SDK type.
  • ArmState<N> contains joint: JointState<N>, a flange StateView<SpatialSample>, optional TCP/stiffness views, and optional load data. Each StateView<T> preserves meas, cmd and des.
  • LoadState describes payload mass, center of mass and inertia.

Why Not Runtime Enums

Older APIs used runtime enums such as MotionType / ControlType. The current mainline uses typed spaces:

robot.move_to::<JointSpace<6>>([0.0; 6])?;
robot.control_with::<TorqueControl<7>, _>(...)?;

Benefits:

  • The compiler can distinguish joint position and joint torque even if both are [f64; N].
  • A driver can implement each supported space/channel separately.
  • Generic upper layers can state their required capability exactly through trait bounds.