Hierarchy Accelerated Stochastic Collision Detection

摘要: In this paper we present a new framework for col- lision and self-collision detection highly de- formable objects such as cloth. It permits to effi- ciently trade off accuracy speed by combining two different collision approaches. We use newly developed stochastic method, where close features of the are found track- ing randomly selected pairs geometric primi- tives, hierarchy discrete oriented polytopes (DOPs). This bounding volume (BVH) is used narrow regions random generated, therefore fewer samples nec- essary. Additionally cost in each time step BVH can be greatly reduced compared pure BVH-approaches using lazy update. For example cloth simulation it experimentally shown that not necessary respond all collisions maintain stable simu- lation. Hence, tuning computation devoted possible yields faster simulations.