Solving Constraint Optimization Problems in Anytime Contexts

Samir Loudni, Patrice Boizumault

Abstract: This paper presents a new hybrid method for solving constraint optimization problems in anytime contexts. Discrete optimization problems are modelled as Valued CSP. Our method (VNS/LDS+CP) combines a Variable Neighborhood Search and Limited Discrepancy Search with Constraint Propagation to efficiently guide the search. Experiments on the CELAR benchmarks demonstrate significant improvements over other competing methods. VNS/LDS+CP has been successfully applied to solve a real-life anytime resource allocation problem in computer networks.