Multi-population Genetic Algorithms with Space Partition for Multi-objective Optimization Problems

It is difficult for the existing multi-population genetic algorithms with space partition to be successfully applied to multi-objective optimization problems. Multi-population genetic algorithms with space partition for multi-objective optimization problems are designed in this paper in allusion to the characteristics of multi-objective optimization problems. A complicated optimization problem is converted into several simple optimization problems. Crossover operator for an intra-population evolution has a direction by using information from the super individual archive. The frequency of updating the super individual archive decreases via pre-selecting optimal solutions submitted to the super individual archive. The search scope of a population is expanded via an inter-population evolution. It is shown from analysis that the computational complexity of the algorithm in this paper decreases evidently. The efficiency of the algorithm in this paper is validated through a complicated benchmark multi-objective optimization problem.