A Multi-objective Evolutionary Algorithm With Binary Quality Indicator

Usually the optimal solution of a singleobjective optimization problem is unique and definite,and solutions of multi-objective optimization problem are a set because of the between of the objective functionsare irreconcilable.With the development of society,the target dimension of many practical problems not only from two to three,such as vehicle path planning,power system management,workshop scheduling,cloud computing,data mining and so on,therefore,high-dimensional multi-objective optimization problem is a main research direction of multi-objective evolutionary algorithm in recent years.The main research work of this paper as follows:Firstly,thebackground,history and current situation of multi-objective optimization problem are introduced.Then introduce the mathematical model andthe related definition of multi-objective optimization problems.Immediately after,the main method of multi-objective evolutionary algorithm and its advantages and disadvantages are introduced.Secondly,introduces three classical multi-objective evolutionary algorithm in detail,and a brief description of the evaluation performance index of multi-objective optimizationproblems and common test functions.Finally,inspired by the advantages and disadvantages of multi-objective evolutionary algorithm for the mainstream,puts forward A decomposition-based multi-objective evolutionary algorithm with binary quality indicator(IBMOEA/D)and the IBDD algorithm which is improved by IBMOEA/D.IBMOEA/D algorithm can be evolved by two reciprocal archives,the non-dominate solution is preserved in the external archive,using binary quality indicators guide the evolution process,and the working archiveusing decomposition technique,putting the multi-objective optimization problem into a single objective optimization problem by decomposed,and its computational resource distribution through binary quality indicators to guide the work archive update.For the disadvantage of the binary quality index,proposed the improved IBMOEA/D algorithm,namedIBDD algorithm.IBDD algorithm combine the direction vector and binary quality indexesin the external archive to guide the evolution basedon IBMOEA/D algorithm.Comparison of IBMOEA/D,IBDD with NSGA-III and MOEA/D_PBI were carried out using function simulationon the DTLZ series of tests,Comparison of IBMOEA/D with EAG_MOEA/D,NSGA-II and MOEA/D_WS in the simulation of discrete combinatorial optimization problems,calculate the performance index of all algorithms on all tests,and analyze their convergence and distribution,from the experimental results,we can arrival at a conclusion that IBMOEA/D andIBDD algorithm has a better performance in solving all kinds of problems.According to the algorithm based on binary quality indicator has a too strong marginal,we propose a multi-objective optimization method and system based on the binary addition indicatorwith direction vector,called EDV algorithm,it can effectively improve the edge effect of epsilon index,theoretically.Comparison of EDV algorithm with MOEA/D-PBI,MOEA/D-IPBI,NSGA-III MOEA/DD,MOEA/D_DU,? DEA,and IBEA_EPS seven algorithm which are classical algorithm for solving high-dimensional multi-objective optimizationproblem on the DTLZ,DTLZ-1,WFG,WFG-1series of problems,through the experimental data to verify the correctness of the theoretical analysis on the EDV algorithm,generally,it has a better effective to solve all kinds of multi-objective optimization problem than other seven algorithm.So,the algorithm with binary quality indicator and direction vectorcan be better in solving all kinds of problems,on the comprehensive,which has a strong advantages for multi-objective optimization to solve the black box problem.Finally,summarize the whole paper,describe the main work and points out the deficiencies of our work,and last give a brief description of the future research direction.