Solving DisCSPs with Penalty Driven Search

We introduce the Distributed, Penalty-driven Local search algorithm (DisPeL) for solving Distributed Constraint Satisfaction Problems. DisPeL is a novel distributed iterative improvement algorithm which escapes local optima by the use of both temporary and incremental penalties and a tabu-like no-good store. We justify the use of these features and provide empirical results which demonstrate the competitiveness of the algorithm.