An Enhanced Strength Pareto Evolutionary Algorithm Based on Reference Direction for Many-Objective Optimization

Pareto-based optimization algorithms have shown competitiveness for optimization problems with two or three objectives, but they may fail in tackling optimization problems with more than four objectives. The crucial reason of failure of Pareto-based optimization algorithms is the loss of selection pressure as the increase of nondominated solutions. In this paper, we propose an enhanced version of recently suggested strength Pareto evolutionary algorithm based on reference direction, termed E-SPEA/R. For maintaining the diversity, the diversity-first and convergence-second strategy continues to be adopted. To improve the selection pressure among nondominated solutions, the procedure of environment selection is modified in E-SPEA/R. Experimental studies demonstrate that the proposed algorithm is most competitive among compared MOEAs on a series of test problems.