A Novel Many-Objective Optimization Algorithm Based on the Hybrid Angle-Encouragement Decomposition

Abstract

For many-objective optimization problems (MaOPs), the problem of balancing the convergence and the diversity during the search process is often encountered but very challenging due to its vast range of searching objective space. To solve the above problem, we propose a novel many-objective evolutionary algorithm based on the hybrid angle-encouragement decomposition (MOEA/AD-EBI). The proposed MOEA/AD-EBI combines two types of decomposition approaches, i.e., the angle-based decomposition and the encouragement-based boundary intersection decomposition. By coordinating the above two decomposition approaches, MOEA/AD-EBI is expected to effectively achieve a good balance between the convergence and the diversity when solving various kinds of MaOPs. Extensive experiments on some well-known benchmark problems validate the superiority of MOEA/AD-EBI over some state-of-the-art many-objective evolutionary algorithms.