Research On Many Objective Evolutionary Algorithm Based On Multi Preference Of Coevolution

Many-objective optimization Problem is always challenging in the field of multi-objective optimization.The optimization algorithm based on the decision maker's preference is more efficient,and the quality of the solutions are better to meet the needs of decision maker,which has attracted the attention from many scholars at home and abroad.Compared with the traditional multi-objective optimization problem,the performance of the many-objective optimization algorithm is greatly reduced because of the increasing number of problem.The research of the many-objective optimization algorithm based on the decision maker's preference mainly focused on the single preference information.At the same time,most of the existing algorithms based on decision maker's preference need people to give preference information.Without knowing the frontier characteristics of the optimization problem,this will undoubtedly bring great cognitive challenges to the decision maker.This paper studies from three aspects of multi-preference,co-evolution mechanism and implicit preference expression.The PICEA framework can effectively identify the Pareto dominant relationship among individuals,reduce the proportion of non-dominated solutions,and make the individuals approximate toward Pareto frontier using co-evolution mechanism.On the one hand,as for candidate solutions at the same fitness level,the multipreference co-evolution algorithm(PICEA-g)based on the target vector can not distinguish the dominant relationship among each other,resulting the distribution of the obtained solution set unevenly.The individual selection mechanism which is different from the traditional adaptive value assignment method is proposed,and then preference-inspired co-evolutionary algorithm based on the hybrid domination stratege(E-PICEA-g)is proposed.Simulation results show that the proposed algorithm performs better on most test functions.On the other hand,This paper studies the principle of multi-preferences guiding population evolution direction.The influence of different preference region control factors on the convergence and diversity of the algorithm is analyzed.The expression of implicit preference information of decision maker is studied,and proposing a better region selection strategy with strong portability combined with PICEA.This method determines implicit preference information using the ASF function,which focuses on preference region construction.The method using limited computing resources to search for preference regions,improving the overall quality of solution set.