An Efficient Local Search algorithm for Over-Constrained CSPs

Abstract

Abstract In industrial experiences of solving constraint satisfaction problems (CSPs), we may often meet an over-constraint situation, which means that we cannot find a solution that satisfies all the constraints. In such situations, the classical CSP is usually extended to take into account costs, preferences, etc. The new objective is usually to minimize the total constraint violation cost or other criterions. The new problems can be described as valued constraint satisfaction problems (VCSP). VCSP can hardly be resolved by the classical constraint satisfying techniques, and this paper provides an application of the Guided Local Search in solving VCSPs.