基于GPU加速的多物体碰撞检测方法  

Accelerating Collision Detection Based on GPU in the N-Body Environment

作　　者：王磊[1] 王毅刚[1] 

机构地区：[1]杭州电子科技大学图形图像研究所,浙江杭州310018

出　　处：《计算机工程与科学》2009年第12期52-55,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60503056);浙江省科技厅科技计划资助项目(2008C24014)

摘　　要：为了在虚拟环境中更加真实地模拟现实环境中物体的运动,需要在仿真系统中加入碰撞检测模块。现有的碰撞检测算法虽然能够快速检测两个物体是否相交,但在物体数量非常多的场景中,因需要对物体两两进行判断,所以仍无法达到较高的检测速度。利用GPU并行计算的特性,在GPU上增加一个预先剔除的过程,大幅度地快速排除不相交的物体,提高了检测的速度。Collision detection is an essential issue of simulation systems which intends to have more realistic motion of objects in virtual environments. Although the existing algorithms can detect the intersection of two objects quickly, they are not quick enough when the number of objects greatly increases, because collision detection for each pair of objects must be performed at least once. This paper proposes a preprocessing method based on GPU and prune most separate object pairs. It only takes a few milliseconds by using the parallel feature of GPU.

关 键 词：碰撞检测 GPGPU 方向包围盒 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]