In virtual manufacturing environment, operator-system interactions and phys
ical simulation require real time collision detection. A fast and precise a
lgorithm for collision detection is therefore critical to ensure the usabil
ity of virtual reality (VR)-based manufacturing simulation. This paper prop
oses an efficient and precise collision detection algorithm for constructiv
e solid geometry (CSG) represented objects in a virtual environment. It mai
nly takes advantages of the CSG 'divide-and-conquer' paradigm and efficient
distance-aided collision detection for convex bounding volumes. An adaptiv
e selection strategy for optimal bounding volumes of sub-trees of objects i
s used to realize a fast localization of possible collision regions at each
detection step, and decision rules for sub-tree freezing and result evalua
ting are created. Performances of the applications used in a virtual traini
ng system for CNC milling machine manual operations and virtual assembly op
erations have demonstrated the effectiveness of the proposed methodology. (
C) 1999 Elsevier Science B.V, All rights reserved.