|Collision Detection on the GPU - An Implementation of CInDeR|
|Master Thesis Project
Start: December 2004
End: September 2005
|Collision detection in real-time rendering is used in a variety of fields, such as CAD, simulations, robotics and games. It's also one of the main bottlenecks in the real-time rendering loop. This has contributed to the development of GPU-based collision detection algorithms. The goal is to move most of the workload of collision detection from the CPU to the GPU, and thus gain an overall speed-up. We have implemented and examined CInDeR, which is an algorithm for collision detection on the GPU.
One of the main problems with this algorithm is the amount of data that needs to be read back to the CPU. To reduce the data and make the implementation usable, we implemented a mip-map algorithm in Cg. We also looked into occlusion queries and how they can aid in manipulating the result.