Routemark-based Navigation of a Wheelchair

This paper introduces the concept of a psychologically inspired navigation approach for an autonomous mobile system. It presents the implementation of the architecture's basic modules and discusses the higher levels of the control-system that are still under development. An electric wheelchair is used as robotics platform to evaluate the architecture's capabilities. 1 Motivation The representation of spatial knowledge is a research topic in many scientiic elds, e.g. in biology, psychology and robotics. In Ger-many, the Deutsche Forschungsgemeinschaft supports the priority program \Spatial Cogni-tion" that facilitates inter-disciplinary collaboration on this subject. The author's research group is in this priority program and is doing research in the detection of landmarks and routemarks and their use for the navigation of an autonomous mobile system. An architecture with multiple layers has been chosen for the control-system of the autonomous system. In contrast to earlier navigation approaches that have been inspired by biological ndings 4, 5] or the reinforcement learning theory 3], our current approach follows certain models that have been discussed with the psychological working groups in the priority program. It consists of the levels \ba-sic behaviors", \route knowledge" and \survey knowledge". Basic behaviors provide a means to robustly move a mobile system through a dynamic world. Route knowledge describes how to navigate along static routes. Survey knowledge is the information about the spatial relationship between the routes. Figure 1: The wheelchair.