Analogical Representation of RCC-8 for Neighborhood-Based Qualitative Spatial Reasoning

Qualitative representations of spatial knowledge aim to capture the essential properties and relations of the underlying spatial domain. In addition, conceptual neighborhood has been introduced to describe how qualitative spatial relations may change over time. Current qualitative representations mainly use symbolic constraint-based languages that are detached from the underlying domain with the downside that a well-formed sentence is not necessarily consistent. This makes it difficult to design efficient knowledge manipulation techniques that consistently advance a representation with respect to conceptual neighborhood. In this paper we argue for analogical spatial representations that inherently obey domain restrictions and, as a result, are consistent per se. We develop a graph-based analogical representation for RCC-8, the construction of which is based on neighborhood transitions realized by efficient graph transformations. The main benefit of the developed representation is an improved efficiency for neighborhood-based reasoning tasks that need to manipulate spatial knowledge under the side condition of consistency, such as planning or constraint relaxation.