Relaxation of Qualitative Constraint Networks

In this paper, we propose to study the interest of relaxing qualitative constraints networks by using the formalism of discrete Constraint Satisfaction Problem (CSP). This allows us to avoid the introduction of new definitions and properties in the domain of qualitative reasoning. We first propose a general (but incomplete) approach to show the unsatisfiability of qualitative networks, by using a relaxation on any set of relations. Interestingly enough, for some qualitative calculi, the proposed scheme can be extended to determine the satisfiability of any qualitative network, leading to an original, simple and complete method. However, as the efficiency of our approach depends on the chosen relaxation, total relations should be preferred due to their connections with the hardness of constraint networks. We then present some preliminary experimental results, with respect to unsatisfiability, which show some promising improvements on some classes of random qualitative networks.