Extended linked voxel structure for point-to-mesh distance computation and its application to NC collision detection

  • Authors:

  • Steffen Hauth;Yavuz Murtezaoglu;Lars Linsen

  • Affiliations:

  • School of Engineering and Science, Jacobs University, Bremen, Germany;ModuleWorks GmbH, Aachen, Germany;School of Engineering and Science, Jacobs University, Bremen, Germany

  • Venue:
  • Computer-Aided Design
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

When working with milling or polishing robots and large workpieces it is necessary to check not only the milling or polishing tool for collision, but it is also necessary to check the remaining arms of the robot for collision. In most of the cases the arms of the robot do not collide with the workpiece and so applying an existing collision detection algorithm to the arms of the robot slows the process down. In this paper, we present an algorithm for quickly assuring non-collisions, which is especially targeted at collisions of the arms of the robot with a workpiece. The algorithm is based on an extended voxel structure. More precisely, we extend a voxel structure by adding distance values to the corner of the voxels and by linking empty voxels to non-empty voxels to accelerate finding the desired voxel. This ensures that we only need to consider a small subset of the triangles describing the workpiece's surface, namely those triangles that are close to the possible collision area. The triangles within each non-empty voxel are stored in a bsp-tree. For empty voxels, we save information about the distances to the mesh. This setup speeds up the point-to-mesh distance calculation, especially for points close to the mesh. The extra distance information in empty voxels enables a fast distance estimation and hence a fast early collision check.