Novel Degrees of Freedom, Constraints, and Stiffness Formulation for Physically Based Animation

Abstract

I identify and improve upon three distinct components of physically simulated systems with the aim of increasing both robustness and efficiency for the application of computer graphics: A) the degrees of freedom of a system; B) the constraints put on that system; C) and the stiffness that derives from force differentiation and in turn enables implicit integration techniques. These three components come up in many implementations of physics-based simulation in computer animation. From a combination of these components, I explore four novel ideas implemented and experimented on over the course of my graduate degree. Eulerian-on-Lagrangian Cloth Simulation resolves a longstanding problem of simulating contact-mediated interaction of cloth and sharp geometric features by exploring a combination of all three of our components. Bilateral Staggered Projections for Joints explores the constrained degrees of freedom of articulated rigid bodies in a reduced state to extend the popular Staggered Projects technique into a novel formulation for rapid evaluation of frictional articulated dynamics. Condensation Jacobian with Adaptivity looks at using reduction methods to improve the efficiency of soft body deformations by allowing larger time step in dynamics simulations. Finally, Ldot: Boosting Deformation Performance with Cholesky Extrapolation explores the inner workings of sparse direct solvers to introduce a Cholesky factorization that is linearly extrapolated in time, which can improve the performance when encapsulated inside an iterative nonlinear solver.