Effects of Problem-Specific Local Search Schemes in a Memetic EMO Algorithm

In some studies, local search has been combined with an evolutionary multiobjective optimization (EMO) algorithm to improve the search ability of the EMO algorithm. Hybridization of an EMO algorithm with local search is often referred to as a multiobjective genetic local search (MOGLS). Such a hybrid algorithm is also called a multiobjective memetic algorithm. Multiobjective memetic algorithms have higher search ability than pure EMO algorithms. In most of combinatorial optimization problems, there exist suitable local search schemes for each problem. These suitable local search schemes can be much more efficient than genetic search. In this paper, to make the search more efficient, we develop the problem-specific local search schemes into a simple multiobjective genetic local search (S-MOGLS) algorithm which we proposed in the previous work. We develop local search schemes suited for multiobjective knapsack problems and multiobjective flows hop scheduling problems. We show the effectiveness of the improved local search schemes through computational experiments.