Efficient Collision Detection for Large Models Employing Filter Banks

This article presents a fast collision detection technique for all kinds of rigid bodies. The new technique can be applied to surfaces, polygon soups as well as volumetric models. It has been developed particularly with regard to large models. It uses a filter bank approach in a preprocessing step, that allows object refinements on-the-fly. At the same time, it prevents collision misses and ensures the detection of a collision at every level of detail, independently from the structure and size of the objects to test. In contrast to common filter banks used by wavelet representations our filter bank construction uses the max-plus algebra instead. This guarantees, that each low resolution version of an object encloses all versions of higher resolution.