A Multi-Objective Chaotic Particle Swarm Optimization Algorithm Based on Improved Inertial Weights

In order to enhance the convergence and distribution of multi-objective particle swarm algorithm, an improved multi-objective particle swarm optimization algorithm was proposed. Linear decreasing inertia weight was used to update. The method can improve the deficiency that the algorithm falls into local optimal easily. The improved Logistic mapping was used to increase ergodicity of the particles. The method can expand the search scope. At the same time, the elite archiving mechanism and the mutation probability were introduced to increase the disturbance. The method can improve the local optimal. Compared with the real Pareto front, NSGA-II and MOEAD algorithm, the simulation shows that the algorithm proposed in the paper is effective. Introduction There are many optimization problems in real life, and meanwhile, the optimization problems need to optimize multiple objectives at the same time. In general, the multiple objective functions in the same problem are contradictory and interaction [1] . So the final results are a series of compromise solutions being called Pareto optimal solutions sets or non-dominated solution sets. Because particle swarm optimization (PSO) has the features of less parameter, good optimization performance, and fast convergence speed. It has been widely concerned and been successfully applied in many fields such as function optimization, data mining and wireless sensor network in recent years. Focusing on the research status at home and abroad, the main research results are summarized as improving the updating strategy of external archive, the selection strategy of global guide, combining the particle swarm algorithm with other algorithms and so on. In literature [2], the external particles were used to guide the flight of the population particles, and the optimal solutions in the group search process were stored in a file. The algorithm is easy to fall into the local optimum being used to optimize multi-peak function. In literature [3], ε-domination was introduced to determine the global extremum of particles. The method can improve the diversity and the uniformity of particles distribution. In literature [4], the preference ε-Pareto domination was introduced to short search time and enhance convergence. In literature [5], a dynamic domain strategies and new particle memory strategies were adopted in the algorithm. In literature [6], a multi-objective interactive PSO algorithm was proposed. To maximize the non-dominated solutions with the adaptive grid mechanism, adaptive mutation operation and user decision functions were used in the algorithm. In literature [7], Pareto entropy was proposed to assess population diversity and current evolutionary states. The evolution strategy can be designed to balance the convergence and diversity with the Pareto entropy. In literature [8], a multi-sub-population co-evolution mechanism was proposed. An external archive and elite learning strategies were introduced to improve distribution and convergence of the algorithm. In literature [9], Chaotic sequence was used to initialize population. The population can be more evenly distributed in the decision space with the method. According to the above researches, a chaotic particle swarm multi-objective algorithm based on linear decreasing inertia weight was proposed in this paper. Chaos mapping was used to optimize