Articulated-Body Simulation (Articulation)

Simulating a robot as rigid bodies connected by joints, computing how it moves and what forces it feels.

A robot is built from links (rigid parts) connected into a tree by joints, and articulated-body simulation is the technique for computing how such a system moves. There are two general approaches: treat every link as an independent rigid body and hold them together with constraints (maximal coordinates), where joints can slowly drift apart or separate; or describe the whole robot using only the base pose plus each joint angle (reduced, or generalized, coordinates), where joints can never come apart by construction — Featherstone's 1987 book laid out the efficient algorithms for this approach. Both PhysX's Articulation feature and MuJoCo take the reduced-coordinate route; PhysX's documentation states that this gives zero joint error and lets it handle much larger mass ratios, with computation scaling with the number of degrees of freedom rather than the number of links. The tradeoff is that it only supports tree structures — a closed loop, such as a parallel ankle mechanism, needs extra constraints added on top.

ExampleImporting a Franka arm's URDF into Isaac Sim produces a fixed-base Articulation: 7 revolute joints plus 2 prismatic finger joints, so the simulation state is just the positions and velocities of these 9 joints.

Also called

Multi-rigid-body simulation, Articulation

Related

Rigid-Body Simulation · Articulated Body Algorithm · Generalized Coordinates · Generalized (Reduced) Coordinates vs. Cartesian (Maximal) Coordinates · PhysX · MuJoCo (Multi-Joint dynamics with Contact)

Sources

NVIDIA PhysX 5 Documentation: Articulations
MuJoCo Documentation: Computation
Wikipedia: Featherstone's algorithm