Dissertation > Excellent graduate degree dissertation topics show

The Research on GPU Based Collision Detection Algorithm

Author: ShiQi
Tutor: XiaoDeGui
School: Hunan University
Course: Information and Communication Engineering
Keywords: Collision detection Hierarchical tree Multicore heterogeneous Graphics Processor
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 348
Quote: 4
Read: Download Dissertation

Abstract

Collision detection between two or more objects are used to determine whether to contact or penetrate, computer graphics, virtual reality, computer games, and CAD / CAM and other areas is a very important issue. After years of exploration, the researchers appear a lot more mature to a certain extent collision detection algorithm. However, with the application of three-dimensional model of the increasingly complex and interactive real-time and scene authenticity requirements to improve, to the efficiency of the collision detection algorithm put forward higher requirements. At the same time, the graphics processor Universal rapid development of computing technology, provides a new approach to the problem of real-time collision detection. In-depth analysis of existing collision detection algorithm based on the combination of general-purpose GPU computing platform CUDA design and improved collision detection algorithm from the following three aspects: First, calculate the angle of departure from the GPU Universal, is proposed based on GPU general-purpose computing collision detection algorithm. Mapped to GPU multi-threaded algorithm for an object between all the triangles of the triangle and the other object intersection test to determine whether a collision between two objects; CUDA platform will intersect test calculation to take advantage of the GPU's parallel computing ability. , the complete parallelization collision detection. Secondly, on the basis of the above algorithm, the introduction of a hierarchy tree is to improve the performance of the algorithm, to obtain a GPGPU-based real time collision detection algorithm. Algorithm to traverse a hierarchy of objects for collision detection between two objects, the bounding box tree, the triangle between the bounding box with another object intersection test, quickly removed the two objects clearly disjoint triangles will intersect the computational complexity from O (N2) down to O (N); intersection test then potentially intersecting triangles between two objects, collision detection results. All the intersection test calculations were performed by the GPU multithreaded parallel execution. The algorithm uses the idea of ??\Finally, based on the principle of CUDA programming optimization, the optimized algorithm GPU memory access and thread degree of parallelism. Memory access, effective use of CUDA shared memory mechanism, the global memory reads data from once a thread is reduced to once per block; threads degree of parallelism, computation and GPU computing power of the grid and The dimensions of the block for a reasonable allocation. Experimental results show that the GPGPU-based real-time collision detection algorithm, able to adapt to the collision detection between objects of arbitrary shape, and has a good real-time performance. Results of this research, the theoretical and practical significance of the real-time collision detection between objects, also played a role in promoting the development of GPU Computing.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile