ALEF: An autonomous vehicle which learns basic skills and constructs maps for navigation

Abstract This paper introduces a control system for a mobile robot which provides learning for two essential kinds of knowledge representation: internal world models and internal abstract concepts. Concretely, learning methods are applied to evolve basic skills (goal oriented driving and collision avoidance) and to generate maps of the environment, even if the mobile robot acts in an a-priori unknown environment and no external navigation aids (like beacons) are used. Results of experiments are presented performed with a real mobile robot equipped with simple and inaccurate sensors (a ring of ultrasonic sensors and goniometers on the wheel axles).