Hybrid Multi-objective Particle Swarm Optimization Algorithm Based on Particle Sorting

In order to enhance the effect of multi-objective particle swarm optimization, a hybrid multi-objective particle swarm optimization algorithm based on particles sorting (PSHMOPSO) is proposed. The algorithm improves in three aspects. First of all, the algorithm uses the method of dividing each dimension evenly to generate the position data of the initial particles to let those particles be distributed in different positions of the decision space as much as possible. Then, when the number of particles in the external archive exceeds the maximum, the method to maintain the number of particles is to score the particles on the grid with the highest density and delete the particles with higher scores. Next, for the selection method of the global optimal solutions (gbest), the reverse selection strategy is adopted in the early iteration of the algorithm. The particles in the areas with few particles are preferentially selected as the gbest of particles to reduce the possibility of the particle swarm falling into the local optimal solutions in the early stage of the iteration. In the latter stage, the forward selection strategy is used to increase the search intensity of the particle-dense areas to find local optimal solutions of these areas for improving the accuracy and convergence of the algorithm. Finally, experiments show that the new algorithm has a good performance when processing test functions.