Analysis of Broad-Phase Spatial Partitioning Optimisations in Collision Detection

This paper examines a number of broad-phase optimisations which can be used to improve the performance of collision detection in dynamic simulations. Specific focus is placed on hierarchical spatial partitioning techniques such as the octree, k-d tree and binary space partitioning (BSP) tree. A number of experiments are conducted using this subset of partitioning methods in order to evaluate their comparative performances in a controlled simulation. A generic structure is created and described which is able to implement each of the tree-based partitioning methods as well as a number of variations on some of the methods. This structure is generic in that the partitioning method implemented depends only on the initial hyperplanes passed to it as a parameter. The results of these tests are analysed in order to identify factors which may affect the extent to which each scheme optimises the collision detection process. Factors identified include allowable tree depth and branching factor, the choice of new partitioners in the recursion process, and the application domain of the partitioning scheme. CR Categories: I.3.15 [Computer Graphics]: Computational Geometry and Object Modeling—Object Hierarchies