The Research On Orthogonal Design-based Dynamic Multi-Objective Optimization Algorithm

In our daily life,there are many decision-making problems.Some of the problems are static optimization problems(SOPs);the others are dynamic optimization problems(DOPs).Nowadays,researchers have got a lot of achievements on SOPs,but there are still seldom classical dynamic multi-objective optimization algorithms.In addition,DOPs are one of the basic issues commonly in real life.It would speed up the development of this research field if these problems can be solved.In recent years,researchers devote themselves to analyzing and solving this kind of problems.The objectives vectors and/or decision vectors of DOPs are usually correlation with the environment or the time.Hence,the Pareto optimal solution set will be changed over time.It is difficult for researchers to design an algorithm which can be used to solve all this kind of problems.This paper presents an Orthogonal Design-based Dynamic Multi-Objective Optimization Algorithm.This algorithm makes use of the historical optimal solution,and then predicts a new population by considering the properties of continuous DOPs when an environmental change is detected.At the end of the paper,some contrast experiments are carried out and the results proves the effectiveness of the algorithm.The main work of the paper is as follows:(1)A survey on DOPs and DOAs.Brief introductions are made on the background and significance of the DOPs.And the detailed definitions of the SOPs and DOPs are described.The related optimization algorithms are reviewed.(2)Orthogonal design-based DMOA.There are five modules in our algorithm.And the function of these modules is shown as follows:the orthogonal design-based multi-objective optimization algorithm module is used to evolve the population when the environment is not changed.The memory module is used to store useful information from the past that will be used in future.Anticipation module is used to manage the information provided by the two predictors and decides when to use it.Autoregression module is used to predict a new population and the environment detection module is used to find when the environment changs.(3)Experimental analysis.This paper proposes an orthogonal design-based dynamic multi-objective optimization algorithm(ODMOA).Compared the optimization results of ODMOA with the results of the feed-forward prediction Strategy(FPS),DNSGA-II-A and RIS,the performance of ODMOA is much better than the other three algorithms.What's more,the convergence of the solutions is better.