Research Of Reference Points Based Many-Objective Evolutionary Algorithm And Application To Satellite Constellation Design Problem

Reference points based evolutionary algorithm employs a well-distributed reference point set as search direction,to improve its performance on the diversity and reduce difficulty for solving many-objective optimization problems.Although reference points based evolutionary algorithms have apparent advantage on the many-objective optimization problems,they still have the defects and shortages in both the development of the algorithm design and application,according to No Free Lunch Theorem.This dissertation focuses on this class of algorithms,and conducts a systematic study on“how to balance the convergence and diversity in the objective space” and“how to handle effectively both irregular and regular problem”,and promote their applications to the satellite constellation design problem.(1)how to balance the convergence and diversity in the objective spaceMulti-objective evolutionary algorithm is to find a set of nondominated solutions which they are approximated Pareto front and uniform distribution over the Pareto front.For many-objective optimization problem,with the number of objectives increasing,Pareto domination becomes invalid,which leads to a number of nondominated solutions in the population and losing pressure toward the Pareto front.In the existing studies,some algorithms have better convergence,while they don't maintain better diversity.Therefore,how to balance the convergence and diversity in the objective space needs be studied further.(2)how to handle effectively both irregular and regular problemDue to well-distrubted reference points in the reference points based multi-objective evolutionary algorithm,they show great advantage on the problems with regular Pareto front.However,when they are used to solve irregular problem,they don't obtain better solution set.To address the issue,some adaptive reference point method is proposed.Nevertheless,these methods don't effectively handle regular problem.Therefore,how to handle effectively both irregular and regular problem is an issue that is worth studying.(3)Satellite constellation design problemSatellite constellation design problem is a representative multi-objective optimization problem.Design of area coverage constellation needs satisfy many indicatiors.However,many studies select one or two indicators as objective function.Therefore,that reference points based multi-objective evolutionary algorithm is used to solve satellite constellation design problem is a significant study.Based on the above issues,the major work and contribution in this dissertation include:(1)For how to balance the convergence and diversity in the objective space,a novel indicator and reference points co-guided evolutionary algorithm is proposed.Indicator can promote convergence,while reference points can maintain better diversity.Therefore,this paper combines them though association operation,to co-guide search.Through experimental verification and analysis,the proposed algorithm can obtain a nondomiated solution set with better convergence and diversity.(2)For insufficient of convergence information in domination relation,an enhanced domination relation is proposed.In order to maintain better diversity,density selection mechanism based on reference points is employed.Based on above content,a new enhanced domination and density selection based evolutionary algorithm is proposed.Through experimental verification and analysis,the proposed algorithm can enhance domination and obtain better performance on the regular problems.(3)Reference points based evolutionary algorithm can obtain better performance on the regular problem,while on the irregular problem they don't show good behavior.Adaptive reference points based can obtain good performance on the irregular problem,while on the regular problem they don't show good behavior.For the issue,an adaptive reference points based evolutionary algorithm is proposed.Entropy is used to control reference point adjustment.To maintain better diversity of reference points,cosine similarity based adjustment strategy is designed.Through experimental verification and analysis,the proposed algorithm can availably handle both irregular and regular problem.(4)Satellite constellation design problem is a representative multi-objective optimization problem.In this paper,reference points based evolutionary algoirthm is employed to address satellite constellation design problem.In order to obtain better constellation,four indicators are selected as objective function.Through experimental verification and analysis,results show that reference points based evolutionary algoirhtm can achieve satisfied demand constellation.