Multicriteria Optimization in CSPs : Foundations and Distributed Solving Approach

In Constraint Satisfaction and Optimization problems ranging from design engineering to economics, there are often multiple design criteria or cost function that govern the decision whereas, the user needs to be provided with a set of solutions which are the best for all the points of view. In this paper we define a new formalism for multicriteria optimization in constraint satisfaction problems “CSPs” and a multi-agent model solving problems in this setting. This approach separately optimizes different criteria in a distributed way by considering them as cooperative agents trying to reach all the non-dominated solutions. It exploits distributed problems solving together with nogood exchange and negotiation to enhance the overall problem-solving effort. The effectiveness of the approach is discussed on randomly generated examples.