Interleaved Asynchronous Arc Consistency in Distributed Constraint Networks

Distributed Constraint Satisfaction Problem (DisCSP) is an area of research in multi-agent systems. In recent years, several distributed constraint algorithms, which perform a depth-first search in a bottom up manner according to pseudo-trees [21], [22], were proposed. In this paper, we present a new asynchronous algorithm that makes use of the problem structure when possible for solving DisCSPs. The algorithm, Interleaved Asynchronous Arc Consistency (ILAAC), is based on the AMAC algorithm [6] and is performed on a pseudo-tree ordering of the constraint graph. The algorithm needs only a polynomial space complexity. This allows to find solutions more efficiently. The experimental results show clearly the usefulness of constraint propagation technique combined with pseudo-tree re-arrangement either for random problems or for distributed graph coloring in terms of communication cost and computation effort.