Probabilistic Region Connection Calculus

We present a novel probabilistic model and specification language for spatial relations. Qualitative spatial logics such as RCC are used for representation and reasoning about physical entities. Our probabilistic RCC semantics enables a more expressive representation of spatial relations. We observe that reasoning in this new framework can be hard. We address this difficulty by using a factored representation based on Markov Random Fields. We formally present the syntax and semantics of a probabilistic RCC. We then use Markov Random Fields to represent our models compactly. Using this representation, we show a basic algorithm for answering queries about the probability of a relation to hold between two entities. Finally, we illustrate the effectiveness of the new approach experimentally over a small set of examples.