A New Evolutionary Algorithm Based on Interval index and Reference Vector for Many-objective Optimization

When solving multi-objective problems, domination-based and decomposition-based multi-objective evolutionary algorithms are the most widely used and have achieved good results. However, the effectiveness of these algorithms decreases when the optimization problem has a complicated Pareto front or the number of objectives increases. To better solve many-objective optimization problems, this paper combines the merits of dominance and decomposition and proposes an optimization algorithm that employs Interval index and reference vector guidance. The algorithm utilizes the proposed Interval index in environment selection to balance the convergence and diversity of populations. A reference vector optimization strategy based on population clustering is designed to adjust the reference vectors to fit different problems. In this paper, the proposed algorithm undergoes thorough evaluation across 72 cases of 18 benchmark problems and is experimentally contrasted against four advanced optimization algorithms. Experimental results establish that the algorithm proposed in this paper has good competitiveness and superiority in solving some of the many-objective problems.