GPU-Accelerated Collision Detection Using Spatial Hashing

In recent years,cloth simulation systems become increasingly important to commercial ap-plications such as animations,games,and virtual fitting.The cloth is completely flexible object,In the simulation process,it will fold,self-entanglement and so on.Therefore,the collision detection phase of the simulation process is very time-consuming is one of the major bottlenecks for simu-lation system performance.In this paper,this paper researches on this difficult problem.It designs and implements a collision detection algorithm based on GPU(Graphics Processing Unit).The algorithm solves the bottleneck of traditional spatial hashing by appling self-collision culling.At the same time,the memory bottleneck of the BVH(Bounding Volume Hierarchy)solution based on the BVTT(Bounding Volune Test Tree)front is solved by taking advantage of the low memory consumption of the spatial hash algorithm.In this paper,the main work summarize as follows:1.This paper designs and implements a uniform grid spatial hashing collision detection algo-rithm on GPU.It combines normal cone culling test tree with uniform grid spatial hashing collision detection algorithm.The algorithm reduces the computational complexity of broad phase and narrow phase in the traditional spatial hashing algorithm and improves the perfor-mance of the collision detection phase.2.At the GPU end,this paper designs a hierarchical grid spatial hashing algorithm.In a uniform grid spatial hashing algorithm,the grid size is a heuristic parameter that needs to be adjusted before simulation.In order to make the collision algorithm adaptive to the specific collision scene.This paper extends the original uniform grid algorithm to the hierarchical grid.3.This paper optimizes the collision detection pipeline in cloth simulation system.In the iter-ative collision detection and response process,it takes advantage of the spatial continuity of the cloth mesh to further to improve the collision detection performance.4.This paper implementes the collision detection algorithm by Visual Studio 2013 on Win-dows lO.On the NVIDIA GTX 1080 graphics cards,it tests several high resolution scenes,in contrast to CAMA system,this paper gets 5to6 times speedup.