Topological Mapping with Sensing-Limited Robots

Most mobile robot mapping and exploration research makes use of longrange, powerful sensors such as laser rangefinders to create maps. In this paper we take a different approach, creating maps using robots with limited sensing capabilities, most notably in their sensor range. Our prototype robots use only five infrared range sensors with a maximum range of 80 cm; in general, these robots cannot “see” both sides of a hallway. We present an algorithm for such a robot to build a topological map in the presence of sensing and odometry error. In doing so, we develop a paradigm to extend topological mapping to open spaces, long considered a deficiency of topological mapping; we also introduce an evidential approach to the problem of “closing the loop.”