Contingency-Tolerant Robot Motion Planning and Control

Abstract

We address the problem of robot motion planning and control in a partially known environment. Examples of this type of environment include shop-floors, office buildings, con- stri~ctiori sites, and clean rooms. In such environments, the sh;?pcs antl the locations of the largest objects are known in ad- vance. But there are other objects whose locations are changing so often that the robot cannot realistically keep track of them. In order for a robot to operate successfully in this type of en- vironment, it must be tolerant to contingencies - i.e., it must be ablc to efliciently deal with unexpected obstacles while ex- ecu ting planned motions. A contingency-tolerant motion plan- ning and control system is presented in this paper. It combines a "lesser-commitment" planner with an "intelligent" controller. Tlic planner produces a set of paths, called a "channel", rather than a single path, in order to let the controller have more free- dom of clioicc. The controller exploits this freedom by applying a potential field method. We have implemented this system and expcrimcnted with it, using both a computer simulated mobile robot and a real one. Acknowledgements: This research was funded by DARPA con- tract DAAA21-89-CO002 (Army), CIS (Center for Integrated Sys- tcms), CIFE (Center for Integrated Facility Engineering), and Digital Gquipnic~it Corporation.