Closest distance searching by GPU-based massive parallel computation

A simulation system for real-time closest distance searching is developed using GPU-based massive parallel computation technology. The system can be used for minimum distance detection between two three-dimensional (3-D) objects in a virtual environment. First, two 3-D models are converted into vertex/mesh format; then, CUDA-based GPU parallel computation is performed for the nearest neighbor searching, in order to find the closest separation Euler-distance of two objects; at last, Unity3D game engine is used as a testing platform for visualization. Experimental results indicate that the system not only works for arbitrary complex objects (i.e., non-convex object), but also has good real-time capability and high accuracy. Compared with traditional brute force algorithm, our parallel searching is about 50 times faster.