Mobile robot self-localization in large-scale environments

This paper presents a new approach for the absolute self-localization of a mobile robot in structured large-scale environments. The requirements with regard to both, the necessary a-priori knowledge as well as the sensor equipment, are low. The algorithm scales up very well, due to a hybrid representation of the environment that augments a topological map with metric information. As a consequence, the method is especially suited for usage in large-scale service robotics applications. As an example for a future application, the so-called navigation assistant of the Bremen autonomous wheelchair "Rolland" is discussed. The self-localization results presented stem from experiments with the wheelchair on a 2,176 m long indoor and outdoor course on the campus of the Universitat Bremen.