Abstract
This paper proposes an idea of using well studied and documented single-objective optimization methods in multiobjective evolutionary algorithms. It develops a hybrid algorithm which combines the multiobjective evolutionary algorithm based on decomposition (MOEA/D) with guided local search (GLS), called MOEA/D-GLS. It needs to optimize multiple single-objective subproblems in a collaborative way by defining neighborhood relationship among them. The neighborhood information and problem-specific knowledge are explicitly utilized during the search. The proposed GLS alternates among subproblems to help escape local Pareto optimal solutions. The experimental results have demonstrated that MOEA/D-GLS outperforms MOEA/D on multiobjective traveling salesman problems.