Environment modeling and localization technique for an autonomous mobile robot

This paper deals with theoretical and experimental aspects of mobile robotics. The theoretical aspect consists of developing an absolute localization technique based on measurements delivered by a rotating sonar. This technique allows one to localize a mobile robot evolving in an indoor environment (laboratory, factory, etc.). An environment model based on ultrasonic range readings is built (sensor based model) and then compared with the reference model, which is a priori known, using clustering techniques. Identification of specific elements in the sensor model (walls and corners) allows absolute localization of the robot. This technique has been successfully implemented on a built mobile robot with simple sensors and a simple data processing architecture. Experiments and their results are also presented.