Sonar based position estimation system for an autonomous mobile robot operating in an unknown environment

The paper presents a new method of position estimation for robot navigation in a completely unknown environment. The method is different from conventional ones in that it does not need any kind of a priori reference model or man-made landmarks. A series of local maps is built and updated from sonar data while the robot is exploring the unknown environment. Among the constructed local maps, the robot autonomously selects the ones with distinctive features and memorizes them as reference landmarks. An orientation clustering method is developed which enables the robot to extract the features of the map. The maps selected in such a way are then used to estimate the position and orientation of the robot while undertaking the given task. In doing so correspondence indices are defined to determine the corresponding reference map to the current local one among the numerous stored maps. The two maps are matched so as to minimize the discrepancy between them, thus enabling one to estimate the position and orientation of the robot. The usefulness of all these approaches is illustrated with the results produced by a real robot equipped with ultrasonic sensors.