Indicator and reference points co-guided evolutionary algorithm for many-objective optimization problems

A key issue for many-objective optimization is how to balance both convergence and diversity. In this paper, we propose indicator and reference points co-guided evolutionary algorithm, called IREA, to solve many-objective optimization. Indicator Iɛ+ can promote good performance on convergence, while reference points can maintain good performance on diversity. Thus, we innovatively combine them through association operator. Association operator first assigns solutions in population to a reference point. Solutions associated with the same reference point constitute a cluster. Then, new population is updated by solutions selected layer by layer from each cluster based on indicator. In addition, to produce better offspring, a binary tournament mating selection is adopted. Finally, the proposed algorithm is compared with six state-of-the-art algorithms on the two well-known test problems. Experimental results indicate that the proposed algorithm can achieve promising performance in terms of generational distance, spacing and Hypervolume metrics. Especially, for the problem with irregular Pareto front, the proposed algorithm also obtains competitive performance.