high performance computing on graphics processing units: hgpu.org

We present an adaptive algorithm for collision detection between rigid and non-rigid polygonal objects by improving particle sensor-based method which is more efficient in handling deformation of objects than bounding volume hierarchies-based method which has to be updated its bounding representations frequently. However, a problem of particle sensor-based is, a number of particles on each object’s surface when object is deformed and a reporting of many contact of exact collision on each object, are not well-defined and calculated. Moreover, it can be slow down in a large scene by checking the distances of every object. Therefore, this research presents an algorithm which solves all those problems by using the sweep and prune like algorithm to prune any noncolliding objects in the large scene instead of checking the distances of every object, using an idea of particle’s emitter to determine an adaptive number of particle sensor when object is deformed, and using a power in stream computation of graphics processing unit (GPU) to report all contacts of exact collision in particle sensor’s responsible area. A detailed description of this algorithm, a time complexity analysis and experiment results showing this algorithm can be conducted in real-time are include in this paper.