Octree-based hierarchical distance maps for collision detection

Abstract

In this paper, we propose a novel hierarchical representation for discretized distance maps commonly used in robotics for path planning and collision detection applications. We augment the well-known octree structure for representing distance maps in a hierarchical manner. Our augmented octree based representation drastically reduces the expensive memory requirement compared to voxel-based distance maps while providing substantially better collision-detection performance than by using unaugmented octrees. Although our main motivation has been collision detection in robotics, similar octree distance maps will have wide applications in many other areas including machine vision, computer graphics and so on.