BROAD PHASE COLLISION DETECTION USING GRAPHICS PROCESSING UNIT

Abstract

Collision detection is a very important component in computer graphics applications.  However, due to its high algorithm complexity, collision detection usually forms a bottleneck in many of these applications causing the simulation performance to deteriorate.  Earlier algorithms for collision detection are sequential in nature.  The multi-core processor technology is seen as an opportunity to reduce and eliminate this bottleneck by parallelizing the collision detection algorithm.  Therefore, this paper implements the sphere bounding volume in the broad phase collision detection using the sequential and parallel approach separately, in order to identify the simulation performance differences between both approaches.  The algorithm used to implement the broad phase collision detection involved the all-pair test where it is based on the comparison of the objects’ bounding volume to determine if collision occurs.  As an extension, this paper utilizes the graphics processing unit to implement the parallel approach.  The implementation of the broad phase parallel collision detection shows improved frame rate for larger number of objects involved; up to 1.2x faster  compared to the sequential implementation.