The integration of methods of Constraint Programming and Multi-Agent-Systems is discussed in this paper. We describe different agent topologies for Constraint Satisfaction Problems and discuss their properties with a special focus on dynamic and distributed settings. This motivates our new architecture for constraint processing with reactive agents. The resulting systems are very flexible and can be used to process dynamic and distributed problems. We define the local behaviors of the agents in this new approach and verify their collective behavior in the context of CSP.
Georg Ringwelski, Richard J. Wallace