Efficient Collision Detection and Physics-Based Deformation for Haptic Simulation with Local Spherical Hash

Abstract

While real time computer graphics rely on a frame rate of 30 iterations per second to fool the eye and render smooth motion transitions, computer haptics deals with the sense of touch, which requires a higher rate of around 1kHz to avoid discontinuities. The use of haptics on interactive applications as surgical simulations or games, for instance, can highly improve the user experience, and the sense of presence. However, its use in complex simulations involving realistic rendering, deformable objects and collision detection requires the development of very performing algorithms. This paper presents an implementation of mass-spring system adapted to CUDA implementation, and a novel method for collision detection in haptic update rate. Important aspects of the port to parallel programming and the GPU architecture are addressed, as for example, strategy and frequency of memory access. A quantitative experiment is also presented to evaluate our methods capability and scalability.