As-rigid-as-possible solid simulation with oriented particles

Abstract We propose a new as-rigid-as-possible approach to the real-time simulation of physics-based deformable models for interactive applications such as computer games. The key observation is that the efficacy of an embedded oriented particle representation and the stability of a variational implicit formulation of the projective dynamics are complementary to each other. We reformulate the variational implicit formulation to deal with an embedded graph of oriented particles. Our new formulation is extremely stable, and our alternating local/global optimization solver is both easy to implement and computationally efficient. Our method can deal with one-dimensional (cable and rod), two-dimensional (shell), and three-dimensional (solid) models in a uniform manner. Experimental results demonstrate that hundreds of deformable models with an extremely large number of polygons can be simulated robustly in real time using thousands of particles.