Projective Constraint Stabilization for a Power Series Forward Dynamics Solver

A power series expression for the forward dynamics of a closed kinematic chain provides an explicit time-step update of the system state. The resulting numerical differential equation solver applies kinematic constraints to the power series terms for acceleration and higher derivatives of motion. Integrating acceleration determines velocity and position time histories that approximate the constraints to a high degree of precision when using a high order of the expansion. When high precision is not required, a lower order achieves shorter computation times, but that condition results in violation of the constraints in the absence of any correction. Projecting the velocities and positions onto the constraint manifold after each time step produces step changes. This paper determines which choices of linear subspace for this projection give step changes that are equal to the residues of truncating the power series solution for the kinematic portion of the problem. The limit of that power series gives position and velocity time histories that approximate the dynamics while giving an exact kinematic solution. Thus projection onto the constraints in this procedure determines sample values of an underlying solution for the motion trajectories, where that underlying solution is continuous in both velocity and position and also satisfies the kinematic constraints at all times. This property is confirmed by numerical simulation of a Clemens constant-velocity coupling.