A Dynamic Organization in Distributed Constraint Satisfaction

We present a novel dynamic organization to solve DC-SP(Distributed Constraint Satisfaction Problem). DCSP provides a formal framework for studying cooperative distributed problem solving[Yokoo 92]. To solve DCSP, we have developed a simple algorithm using iterative improvement. This technique has had great success on certain CSP(Constraint Satisfaction Problems)[Minton 90][Selman 92]. In our algorithm each agent performs iterative improvement and also plural agents can do in parallel. However, one drawback of this technique is the possibility of getting caught in local minima(which are defined specifically in our algorithm). LMO is a technique for escaping from local minima. It is summarized as follows: