Real-time inverse kinematics through constrained dynamics

Abstract

Motion capture is an essential technique for interactive systems, such as distributed virtual reality, video games, and entertainment. Inverse kinematics algorithms are often used to minimize the number of trackers required for motion capture systems. The solving of inverse kinematics problems can be computationally expensive. We introduce a real-time algorithm for inverse kinematics computation, originally from the field of molecular simulation, called SHAKE. We also describe the implementation of the algorithm in our motion capture system for avatar motion generation through constraint dynamics. We demonstrate that the algorithm has advantages over conventional methods with properties of fast convergence, energy stability, and constraint system consistency when adding additional constraints.