An Adaptive Multi-objective Particle Swarm Optimization Algorithm Based On Decomposition

The multi-objective optimization problems exist widely in many scientific research fields.The different goals in a multi-objective problem is mutually exclusive and influential,when designing an algorithm,it is often impossible to make every goal of multi-objectives optimal.Therefore,the objective of multi-objective optimization is to solve the problem.Find out as many solutions as there are nondominated relations(called Pareto optimal solution sets).In this paper,the multi-objective evolutionary algorithm is studied in depth,in the framework of MOPSO/D,the algorithm was improved from two aspects.One is to improve the efficiency of searching Pareto set by improving the speed and position update formula of particle swarm optimization algorithm.At the same time,the inertia weight and acceleration factor of the particle swarm are improved to make it self-adaptively adjusted,which can better balance the global and local search.Second,the grid technology is used to store the optimal solution set,which can effectively maintain the distribution of evolutionary populations.Sexuality is tested by using five classic two-target test functions.The experimental results show the effectiveness of the algorithm.To solve the problem of AMOPSO/D for solving three goals and more objective problems,based on MOPSO/D,a fuzzy control-based multi-objective particle swarm optimization algorithm based on decomposition is proposed.This algorithm adopts fuzzy controller,the diversity and the number of iterations of particle swarms is used as the input of the fuzzy controller.The inertia weight and acceleration factor of the particle swarm are dynamically adjusted through the output of the fuzzy controller.The test experiment is performed through a series of three-target test functions.The experimental results are compared with other classical algorithms,indicating that the algorithm to the three-objective problem shows better performance than other algorithms.Finally,this paper compares MOEA/D,AMOPSO/D and FMOPSO/D on five ZDT series test functions,further illustrate the superiority of the performance of this algorithm.