The Virtual Assembly Technology Based On Parallel Continuous Collision

The collision detection problem, as an aspect to evaluate a virtual assembly system wassuperior or not, has attracted widely attention. The key criterion for judging a collisiondetection algorithm was the real-time performance and the accuracy, as a result, how to usethe parallel architecture of increasingly sophisticated computer hardware to achieve real-timeand accurate collision detection of mass complex scene has become the hot topic of currentresearch.Currently, the continuous collision detection is still a big challenge to be performed inreal time, especially in these complex scenes. Based on the CPU parallel architecture, weproposed a multi-core collision detection algorithm to improve the overall performance ofvirtual assembly system. Specifically, we first introduce a spatial decomposition theory. Andthen present an allocation strategy of dynamic tasks to distribute the collision detection tasksto all kernels equally. Finally, through the analysis of the data, our algorithm can improve theefficiency of collision detection.Compared with the CPU, which hast strong logical computation ability, the GPU hashigher powerful floating-point computation ability, which lead to the advantage in theassembly system. However some errors may occurred during the mathematical modelingprocess, which may cause the fact that the part collision may not be contacted due to theoverlap or nest phenomenon.In order to solve the above problems, we use the powerful computing ability and highdensity performance of the multi-core of modern GPU, to present a parallel collision detectionalgorithm in a real-time GPU acceleration circumstances. The main contents are listed asfollows:First, we proposed a new framework based on the GPU acceleration parallel frameworkfor the virtual assembly system, the continuous collision detection is performed to avoid theomission phenomenon of collision in the process of moving objects.Then, in order to solve the communication bottleneck problem between CPU and GPU,which is caused due to the bandwidth limitations by the data throughput, we propose the CCDmethod based on collision flow optimization method which does not need to frequentlytransmit large data quantity so as to improve the overall performance of collision detectionalgorithm, we also implement our system real time.Finally, our algorithm combines the discrete collision detection algorithm with the continuous collision detection, we also integrate the triangular mesh collision detection(Triangle Intersection Detection, TID) to our current multi-core collision detection algorithm,which is based on the CCD. Based on the hybrid collision detection, our algorithm has theability to avoid the undetected collision which is caused by overlapping each other or having anested phenomenon between parts during the initialization process.