V-Clip: fast and robust polyhedral collision detection

Reviewer: Andrew Timothy Thornton

V-Clip (or Voronoi Clip) is an algorithm, based on the Lin-Canny and GJK algorithms, for determining the closest pair of features between convex polyhedra modeled using boundary representations. By using the topology of the convex polyhedron, the algorithm moves from feature to feature, either until it finds that the polyhedra have intersected, or until it finds the pair of features with the minimum separation distance. The result of this approach is that the algorithm is more robust than Lin-Canny (in particular, in its handling of intersecting polyhedra), and offers better performance than the GJK algorithm. It also offers a general improvement in numerical robustness. Because the algorithm works from a given starting point on each body, it is particularly suited to successive tests of the separation of bodies moving relative to each other. The performance tests provided in the paper are based on different levels of coherence between objects. The algorithm is restricted to convex objects, so nonconvex objects must be decomposed into convex elements, or other approaches such as octrees or bounding boxes may be more appropriate. The V-Clip algorithm could be a useful tool to those involved in animation or other areas where it is necessary to detect the proximity or collision of moving objects.