DisBO-wd: a Distributed Constraint Satisfaction Algorithm for Coarse-Grained Distributed Problems

Abstract

Abstract. We present a distributed iterative improvement algorithm for solving coarse-grained distributed constraint satisfaction problems (DisCSPs). Our algorithm is inspired by the Distributed Breakout for coarse-grained DisCSPs where we introduce a constraint weight decay and a constraint weight learning mechanism in order to escape local optima. We also introduce some randomisation in order to give the search a better chance of finding the right path to a solution. We show that these mechanisms improve the performance of the algorithm considerably and make it competitive with respect to other algorithms.