A Differential Evolution Framework with Two Subpopulations for Handling Multi-objective Optimization Problems

A differential evolution framework with two subpopulations for multi-objective optimization is presented. Based on the strength values of the individuals, the current population is divided into two subpopulations to search the space of solutions effectively. One subpopulation consists of the top individuals of the current population and employs scheme DE/best/1/bin to improve the convergence speed by learning the non-dominated individuals; the other is composed of the rest individuals of the current population and uses scheme DE/rand/2/bin to improve and guarantee the convergence rate. Through validating the two performance metrics on three benchmark multi-objective optimization test problems, the presented algorithm is compared with three state-of-the-art algorithms. Simulation results show that the presented algorithm can obtain better performance. Thereafter, three scalable test problems for evolutionary multi-objective optimization are tested to further demonstrate that the algorithm is feasible and effective.