Controlling a mobile robot using partial representations

Abstract

Multiple partial representations greatly facilitate the construction of complex autonomous robots. Such representations are desirable since they are easy to develop and quick to compute. We describe how to select various representations, how to string them together to perform specific tasks, and how to provide the resulting distributed system with an overall goal. As with the original Subsumption Architecture, our approach allows the incremental development of a system by the accretion of successive performance layers. The application of these techniques is illustrated by the analysis of an existing mobile robot which collects soda cans in an unstructured office environment.