A new hybrid memetic multi-objective optimization algorithm for multi-objective optimization

Abstract

To deal with the multi-objective optimization problems (MOPs), a meta-heuristic based on an improved shuffled frog leaping algorithm (ISFLA) which belongs to memetic evolution is presented. For the MOPs, both diversity maintenance and searching effectiveness are crucial for algorithm evolution. In this work, modified calculation of crowding distance to evaluate the density of a solution, memeplex clustering analyses based on a grid to divide the population, and new selection measure of global best individual are proposed to ensure the diversity of the algorithm. A multi-objective extremal optimization procedure (MEOP) is also introduced and incorporated into ISFLA to enable the algorithm to evolve more effectively. Finally, the experimental tests on thirteen unconstrained MOPs and DTLZ many-objective problems show that the proposed algorithm is flexible to handle MOPs and many-objective problems. The effectiveness and robustness of the proposed algorithm are also analyzed in detail.