The use of qualitative spatial knowledge representation has significant advantages over purely quantitative spatial representations. It allows abstraction from irrelevant details on different levels of granularity and abstraction and allows to generalize large sets of quantitatively different spatial situations into a single description and therefore provides the foundation for more abstract e.g., sensor-independent behavior models. Nevertheless a qualitative spatial knowledge representation is rarely used to describe (complex) behavior for autonomous agents in physically grounded environments. One of the main reasons is the lack of robust methods that support the generation of qualitative spatial descriptions from quantitative sensor input. In this paper we present two approaches that provides support to the generation of static and dynamic qualitative spatial representations.

Andrea Miene, Thomas Wagner

30-35

2