Research On The Adaptive Mechanism Of Multi-objective Evolutionary Algorithm Based On Decomposition

Mops are inevitable conflict problem in our daily life.The research and optimization on algorithms which can solve such problems has been always playing an important role in the field of Intelligent Computing.At present,the multi-objective evolutionary algorithm based on decomposition has become hot topics in the field of evolutionary computation,because of its advantages of easy expansion,high accuracy,fast convergence,especially the low complexity on solving high-dimensional problems.However,in the actual solving process,the MOEA/D uses a fixed scale neighborhood,and the selection process is lacks of overall control.In particular,when dealing with complex and high dimensional problems,the algorithm lacks adaptive adjustment.Therefore,the algorithm can't balance the convergence and distribution,resulting in low evolutionary efficiency and unfairness in the selection process.In order to make the algorithm serve the field of engineering applications better,the research of the adaptive mechanism of MOEA/D has gradually become a hot spot of concern.In this case,in order to improve the adaptive performance of MOEA/D,the judgment mechanism which can reflect the state of population evolution is put forward,by exploring the rule of population evolution.On the basis of this,the research on the adaptive mechanism of the neighborhood setting and selection strategy of MOEA/D is carried out.The specific work is as follows:1.An adaptive neighborhood strategy is proposed.By exploring the evolution law of MOEA/D and the information of individual neighborhood update,an evolutionary potential judgement mechanism based on neighborhood updating capability,which can better respond to population evolution and evolution state,is summarized.Then,based on the judgement mechanism and the convergence and distribution requirements of the algorithm in the evolution process,we propose an adaptive neighborhood strategy based on evolutionary potential judgement mechanism(ANS).Let algorithm set the different neighborhood size according to the different evolutionary states of the population and the individual.Finally,it is proved by experiments that the strategy effectively balances the convergence and distribution of the population in the evolutionary process,and improves the overall performance of the set.2.An adaptive selection strategy is proposed.Using neighborhood updating as the main selection strategy,MOEA/D focus more on local replacement.In order to overcome this flaw of MOEA/D,a new selection strategy based on perfect matching of bipartite graph(KMS)is applied to select the elite solutions from a global perspective by matching the solution with suitable subproblem.Then analyze the advantages of neighborhood updating and KMS.An adaptive selection strategy based on the disorder potential judgment(AS)is proposed in combination with the evolutionary potential judgment mechanism.The experimental results show that the MOEA/D algorithm using AS has better convergence and distribution.It proves that AS can guide the selection process of elite solution effectively.3.Adaptive MOEA/D algorithm is applied to solve complex and high-dimensional problems.Firstly,the ANS and AS strategies are combed and integrated.Then,the two strategies are incorporated into the evolutionary framework of the MOEA/D,and eventually an adaptive MOEA/D(AMOEA/D)is constituted.Through experiments on standard test functions,it is proved that AMOEA/D can search for relatively high-quality front in solving various complex and high-dimensional optimization problems,and it has better stability and robustness.