A Close Neighbor Mobility Method Using Particle Swarm Optimizer For Solving Multimodal Optimization Problems

There are many optimization problems need to find as many global optimal solutions as possible in the real world,such as in a power system^[1],protein structure prediction^[2],data mining^[3,4,5].These problems are commonly known as multimodal optimization problems?MMOPs?.There are two aspects worthy of our attention regarding the optimization of multimodal problems.The first aspect,for most multimodal optimization algorithms,if the required accuracy is high,they fail to find the desired optima even after converging near them.In the second aspect,many algorithms have difficulty exploring all regions,and some of the highest peaks are easily missed.Especially for some spikes,the decision space occupied by their peaks is very small,and it is difficult for particles to explore these peaks.In the field of evolutionary computing,most of the algorithms are based on niche technology for dealing with multimode optimization problems.Therefore,in this paper,we introduce several algorithms with completely different styles to solve this kind of problems,which provide a variety of different ideas for solving multimode problems.MOMMOP and DR?CRA algorithms are used to solve single target multimode problems.MOEA/D?AD and MO?Ring?PSO?SCD are used to solve multiobjective multimode problems.In addition to the MOMMOP algorithm,the core essence of the above algorithms is niche technology.Most of the existing multimode algorithms try to find all the global optimal solutions with high accuracy,but they do not get ideal results.The algorithm proposed in this paper can effectively improve the accuracy of multimodal optimization solution,because multimodal optimization algorithm can not find the exact optimal value.The validity and feasibility of the method are verified by experiments.This paper has developed a particle swarm optimizer using the close neighbor mobility strategy for solving multimodal optimization problems,which can achieve a better balance between exploration and exploitation.Three novel techniques are developed to improve the performance of the algorithm:elite selection mechanism,close neighbor mobility strategy,and modified DE strategy.Based on these three novel techniques,CNMM can achieve a promising performance when dealing with MMOPs,regardless of the accuracy level.The results also show the effectiveness and feasibility of the CNMM for quickly locating more accurate global optimal solution.