Discovering Landmarks for Mobile Robot Navigation

Positioning addresses the problem of determining the position of a mobile robot from sensor data. This paper presents a method with which a mobile robot can automatically discover a set of landmarks used as reference for positioning,and also train artificial neural networks to recognize them. The method is applicable to a large variety of sensors and environments. Extensive experiments with a mobile robot equipped with sonar sensors and a camera, illustrate that landmarks selected by the robot are superior to landmarks carefully selected by human experts. Thus, the new method has two advantages over previous ones: It improves the ability of a robot to position itself, and it eliminates the burden of having to select appropriate landmarks by hand.