Optimal geometric modeler for robot motion planning

Abstract

Planning collision-free trajectories under real-time restrictions is a challenging topic in robotics. In order to reduce computational cost in the collision avoidance process, some authors have proposed different model representations. This article presents an optimal method able to generate automatically geometric models of the objects in a robotic system. For each object, two models are obtained, i.e., the minimum outer and the maximum inner models. Availability of both models allows one to face more successfully robot motion applications. The geometric modeler is focused on the generation of spherically extended polytopes. Each object to model is represented by a set of points taken from its surface. Models are generated through the application of an iterative process based on the Hough transform. When both outer and inner models have been generated, a parameter for evaluating the quality of the models is introduced. This parameter can be used by a rule-based system for increasing the complexity of the model generated and improving, therefore, the accuracy of the representation. © 2000 John Wiley & Sons, Inc.