View article

The Spatial Semantic Hierarchy (SSH) provides a robot-independent ontology and logical theory for building topological maps of large-scale environments online. Existing SSH implementations make very limited use of perceptual information and thus create many candidate maps. Metrical mapping implementations capture detailed knowledge about local small-scale space but do not handle large environments well due to computational limitations and global metrical uncertainty. In this paper, we extend the SSH to utilize better sensory information by incorporating information derived from local metrical models into the large-scale space framework. This new extension of the Spatial Semantic Hierarchy uses local topology obtained from local perceptual models to constrain a global topological map search.