A clustering and vector angle-based adaptive evolutionary algorithm for multi-objective optimization with irregular Pareto fronts

In recent years, multi-objective optimization evolutionary algorithms (MOEAs) have been proven to be effective methods for solving multi-objective optimization problems (MOPs). However, most existing MOEAs that are limited by the shape of the Pareto fronts (PFs) are only suitable for solving a certain type of problem. Therefore, in order to ensure the generality of the algorithm in practical applications and overcome the constraints brought by the shapes of PFs, a new adaptive MOEA (CAVA-MOEA) based on hierarchical clustering and vector angle to solve various MOPs with irregular PFs is proposed in this article. Firstly, a set of adaptively generated clustering centers is used to guide the population to converge quickly in many search directions. Secondly, the vector angle-based selection further exploits the potential of the clustering algorithm, which keeps a good balance between diversity and convergence. The proposed CAVA-MOEA is tested and analyzed on 24 MOPs with regular PFs and 18 MOPs with irregular PFs. The results show that CAVA-MOEA has certain competitive advantages compared with the other six advanced algorithms in solving MOPs with irregular PFs.