Font Size: a A A

The Study On Evolutionary Dynamic Multi-objective Algorithm Based On Central Point Prediction And Subpopulation Guidance

Posted on:2020-04-13Degree:MasterType:Thesis
Country:ChinaCandidate:J W ZhouFull Text:PDF
GTID:2428330578460317Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Dynamic multi-objective optimization problems exist widely in the fields of science and engineering(including air traffic system,control and optimal scheduling,industrial design and robot navigation).Therefore,it is of great economic and social significance to study dynamic multi-objective optimization problems.Dynamic multi-objective optimization problems are more challenging than multi-objective optimization problems,optimization goals not only have multiple dimensions and conflict with each other,but also interfered with factors such as time.Quite literally,the dynamic multi-objective optimization problems refer to the dynamic change of the objective function to be optimized.It is very difficult or even impossible to solve such problems by traditional optimization methods.Evolutionary algorithm is a global optimization algorithm based on population search,which simulates biological evolution and has been widely used.Over the years,the research of evolutionary multi-objective optimization has made good progress.Evolutionary algorithms have achieved good results in solving static multi-objective optimization problems and dynamic single-objective optimization problems.Therefore,it is of great practical and theoretical significance to further study the application of evolutionary algorithm in solving dynamic multi-objective optimization problems.To solve the dynamic multi-objective optimization problem,the main method is to use the prediction strategy to quickly respond to changes in the environment.However,when the problem to be optimized is complex or unpredictable,the prediction strategy can not solve this kind of problem well.This paper proposes an algorithm based on predictive and autonomously guided hybrid strategies that responds to changes in the environment by generating a new population.According to the location of the historical population,a part of the individual is generated by a prediction strategy,which comes at the cost of loss prediction accuracy for the purpose of rapid response.Another part of the individual is generated by an autonomously guided strategy,a subpopulation produced by several generations of current population evolution which directs the current population to evolve toward a better region.The hybrid strategy not only gives full play to the advantages of the prediction strategy in dealing with dynamic multi-objective optimization problem,but also compensates for the deficiencies of the prediction strategy through the sub-population autonomous guidance strategy.In this paper,the proposed algorithm and the other four algorithms are compared and analyzed experimentally.The experimental results show that the proposed algorithm is more competitive on most test problems.
Keywords/Search Tags:dynamic multi-objective optimization, evolutionary algorithms, prediction strategy, sub-population autonomous guidance, center-point prediction
PDF Full Text Request
Related items