A projective method for collision detection of robot arms moving among obstacles

Abstract

The authors present a projective method for collision detection for robot arms moving among obstacles. A mixed model for describing objects is suggested. Procedures for intersection checking among primitive shapes are developed that are based on the idea that two objects have no intersection if one can find a particular projection direction such that the images of the two objects have no intersection on the projection plane. It is noted that not all of these algorithms provide necessary and sufficient conditions for collision detection. However, none of them requires more than two projections to determine whether a collision is likely to occur. Furthermore, images of the objects after projection are simple shapes such as circles and polygons. Thus, the checking procedures are simple and efficient for online implementation. The method can also be applied for planning paths for multiarm robots. Simulation results are included to demonstrate the applicability of the algorithms.<>