Orientation in Extended Position-Based Dynamics: Application to Rigid Bodies and Cosserat Rods

Extended Position-Based Dynamics has long handled translation well, but rotation has been the awkward part because 3D orientation lives on a nonlinear manifold. A new Lie-theoretic formulation tackles that directly, giving XPBD a unified way to express rotations, constraints, interpolation, and gradients without the usual ad hoc machinery.

The biggest headline for rigid bodies is a set of explicit constraint formulations and gradients that dramatically improve dynamic consistency in constrained XPBD simulations. The measured gain is over 10,000x versus the current state of the art, which is the kind of jump that can matter when you’re trying to keep stacks, joints, and articulated props stable under heavy constraint pressure.

The same framework also extends to finite-element Cosserat rods by interpolating nodal rotations on-manifold. In practice, linear finite elements beat the common chain-of-rigid-bodies approach, while higher-order basis functions produce smoother results and faster convergence. That makes the method interesting for ropes, cables, tentacles, hair-like systems, and other deformables where rotation fidelity matters as much as position.

For game teams, the practical takeaway is that XPBD doesn’t have to treat orientation as a special-case hack anymore. If this approach holds up in production, it could simplify solver code, improve robustness under large deformations and contact, and reduce the tuning pain that often comes with rotational constraints.