CSP-104: Constraint Propagation for Loose Constraint Graphs

In this paper we investigate whether we can improve propagationbased finite domain constraint solving by making use of the constraint graph to choose propagators to execute in a better order. If the constraint graph is not too densely connected we can build an underlying tree of bi-connected components, and use this to order the choice of propagator. As search progresses forward, the constraint graph becomes less and less strongly connected, so if we can determine the bi-connected components of the dynamically shrinking constraint graph we have more scope of making use of the technique. Our experiments show that there exist problems where handling biconnected components can substantially improve the propagation performance.