IRI-9504138 Kuipers, Benjamin University of Texas-Austin $93,501-12 mos An Ontological Hierarchy for Spatial Knowledge Spatial knowledge is an important and ubiquitous kind of commonsense knowledge. it also raises a critical issue facing computational models of mind: the relationship between behavior in the continuous world and inference with discrete symbolic representations. in this project, a hierarchy of representations for spatial knowledge called the Spatial Semantic Hierarchy (SSH) developed in a previous NSF grant is being formalized so that it can be more widely useful. The formalization will clarify precisely when and how different descriptive ontologies and inference methods can interact effectively. It will also identify the minimal preconditions on sensors, effectors, and environment required for the spatial model to apply and identify the minimal set of targets for machine learning and discovery required to learn an instance of the hierarchical spatial model from experience. It will be evaluated both mathematically and experimentally on simulated and physical robots. The results should be theoretically important to the foundations of artificial intelligence and practically important to intelligent robotics applications.
