Automatic generation of an optimal guide map for a mobile robot

This paper describes an approach to automatic map generation for an autonomous vehicle. In the first phase, the vehicle is driven by a human through a corridor. During this phase, a vision system obtains many images with a TV camera mounted on the vehicle, and extracts features such as doors or staircases as candidates of landmarks. In the second phase, the vehicle system examines the extracted features and selects only useful features for guiding the vehicle. These features constitute a map for autonomous driving. Preliminary experiments are performed for an autonomous vehicle which moves in our building.