Autonomous mobile robots emerge as rather universal experimental tools for exploring percepual knowledge, its relation to knowledge representation, inference drawing about their environments, and the use of these inferences to act in a physical world. As autonomous mobile robots must act in both, the real physical environments and their formal representations, the relation between the ‘real world’ and its static and dynamic representations are of particular interest. Processes of spatial reasoning and navigation can be investigated all the way from the perception of the environment to performing actions in the environment. Also, autonomous mobile robots can be used to test and explore theories of spatial cognition developed in cognitive psychology.

I will report about two projects carried out in the framework of the German Spatial Cognition priority program funded by the Deutsche Forschungsgemeinschaft. In this program, sixteen research teams consisting of computer scientists, cognitive psychologists, geographers, and biologists cooperate on the investigation of spatial cognition - real and abstract, human and machine. In one of the projects (the ‘aspect map’ project) we investigate integrated spatial and non-spatial forms of representing spatial knowledge (as in geographic maps); in the other project (the ‘spatial inference’ project) we develop and study methods and algorithms for qualitative spatial reasoning, i.e. calculating spatial relationships without the use of numbers. In a cooperation between both projects, we study the use of schematic maps for robot instruction, combining map representations with qualitative spatial reasoning. We believe that this is an interesting and useful paradigm for studying multimodal human interaction, as maps integrate analogical spatial knowledge and propositional linguistic knowledge.