Navigation for an Intelligent Mobile Robot

Abstract

This paper describes a system which performs task-oriented navigation for an intelligent mobile robot. Global path planning in this system is based on a pre-learned model of the robot's domain. The pre-learned model is divided into convex regions using a new "maximal-area" convex decomposition algorithm. A network of convex region entry-poiDts, called "adits", provides the basis for planning a global path as a sequence of straight line movements. This navigation system is based on a dynamically maintained model of the local environment, called the "Composite Local Model", which integrates information from different sensors and different views, as well as from the pre-learned model of the robot's domain. Local straight line movements are planned and monitored using this local model. The estimated position of the robot is corrected by the difference in position between observed sensor signals and the corresponding symbols in the local model. This system is useful for navigation in a finite, pre-learned domain such as a house, office, or factory.