Efficient Conservative Collision Detection for Populated Virtual Worlds

Abstract

Large virtual worlds, with considerable level of detail are starting to emerge everywhere, from large areas of actual cities to archaeological detailed reconstructions of sites. Populating a virtual world adds an extra touch to the visualization of these worlds, but unfortunately it also brings an extra burden to the system. Several tasks are required when adding animated characters to a virtual world, such as collision detection, path planning and other AI algorithms, rendering of dynamic geometry, amongst others. In here a method for efficient and scalable conservative collision detection, that is able to deal with large scenes and thousands of avatars, is presented. This method does not perform exact collision detection, hence it is conservative. The method is suitable as a basis for path planning algorithms and other AI algorithms where an avatar is often regarded as something that can be bounded by a cylinder, or a box. The algorithm is capable of dealing with arbitrarily complex 3D worlds, and does not require any a priori knowledge of the geometry.