AN ENHANCED TOPOLOGICAL MAP FOR EFFICIENT AND RELIABLE MOBILE ROBOT NAVIGATION WITH IMPRECISE SENSORS

An enhanced topological mapping system for efficient and reliable navi gation is presented. The map has a topological framework and some addi tional features. Firstly, it utilizes such rough metrical information as the length and orientation of the links. Secondly, it provides a re liable localization algorithm with which the robot first Ends the inte rval describing the robot's probable location by estimating the projec ted traveled distance using dead reckoning and then fine-tunes the est imation using landmark detection modules. Finally, it provides a plann ing algorithm with which the robot's path is chosen so that the robot reaches the goal location as fast as possible without losing its way d espite using such imprecise sensors as ultrasonic range finders. We ha ve implemented and tested the proposed mapping system both on a simula tor and a real mobile robot, the CAIR-2. This paper also describes lan dmark detection modules that utilize ultrasonic range finders. Althoug h landmark detection modules are too simple and imprecise to estimate position by themselves, these experiments show that the proposed mappi ng system can reliably guide robots. (C) 1998 Elsevier Science Ltd. Al l rights reserved.