Collision Detection between a Complex Solid and a Particle Cloud assisted by Programmable GPU

Abstract

In this work the problem of collision detection (CD) between a Complex Solid and a Particle cloud with autonomous movement is studied. In order to do this, some algorithms and data have been adapted to suit new extensions of the new generations of programmable graphics cards. These types of graphics cards allow more flexible programming in order to solve problems not-related to visualization process. We use a representation based on simplicial coverings as well as a structure named Tetra-Tree in order to represent and classify the simplices of complex objects. With this type of representation the operations carried out in CD are more robust and efficient than those used in classic algorithms, so it is not necessary to decompose the complex object into more simple pieces. We also propose some implementation alternatives and give a study of their performance.