Computer Graphics Laboratory
Image-space techniques have shown to be very efficient for collision detection in dynamic simulation and animation environments. This paper proposes a new image-space technique for efficient collision detection of arbitrarily shaped, water-tight objects. In contrast to existing approaches that do not consider self-collisions, our approach combines the image-space object representation with information on face orientation to overcome this limitation.
While image-space techniques are commonly implemented on graphics hardware, software solutions have been neglected so far. In this paper, the performance of two GPU-based implementations and one CPU-based implementation of the proposed collision detection algorithm are compared. Results suggest, that graphics hardware accelerates collision detection in geometrically complex environments, while the CPU-based implementation provides more flexibility and better performance in case of small environments.