A tabu-based exploratory evolutionary algorithm for multiobjective optimization

Abstract

This paper presents an exploratory multiobjective evolutionary algorithm (EMOEA) that integrates the features of tabu search and evolutionary algorithm for multiobjective (MO) optimization. The method incorporates the tabu restriction in individual examination and preservation in order to maintain the search diversity in evolutionary MO optimization, which subsequently helps to prevent the search from trapping in local optima as well as to promote the evolution towards the global trade-offs concurrently. In addition, a new lateral interference is presented in the paper to distribute nondominated individuals along the discovered Pareto-front uniformly. Unlike many niching or sharing methods, the lateral interference can be performed without the need of parameter settings and can be flexibly applied in either the parameter or objective domain. The features of the proposed algorithm are examined based upon three benchmark problems. Experimental results show that EMOEA performs well in searching and distributing nondominated solutions along the trade-offs uniformly, and offers a competitive behavior to escape from local optima in a noisy environment.