Emergent Constraint Satisfaction Through Multi-Agent Coordinated Interaction

We present a methodology, called Constraint Partition and Coordinated Reaction (CP&CR), for distributed constraint satisfaction based on partitioning the set of constraints into subsets of di erent constraint types. Associated with each constraint type is a set of specialized agents, each of which is responsible for enforcing constraints of the speci ed type for the set of variables under its jurisdiction. Variable instantiation is the joint responsibility of a set of agents, each of which has a di erent perspective on the instantiation according to a particular constraint type and can revise the instantiation in response to violations of the speci c constraint type. The nal solution emerges through incremental local revisions of an initial, possibly inconsistent, instantiation of all variables. Solution revision is the result of coordinated local reaction of the specialized constraint agents. We have applied the methodology to job shop scheduling, an NP-complete constraint satisfaction problem. Utility of di erent types of coordination information in CP&CR was investigated. In addition, experimental results on a benchmark suite of problems show that CP&CR performed considerably well as compared to other centralized search scheduling techniques, in both computational cost and number of problems solved.