Research On A Parallel Many-Objective Evolutionary Algorithm For Large-Scale Optimization And Its Application

Nowadays,5G has attracted more and more people's attention.There is no doubt that the coming of 5G era can greatly improve people's quality of life.But 5G network has not been fully deployed,therefore,it is meaningful to study how to solve the problem of 5G network deployment optimization.At present,most researches on the optimization of 5G network edge device deployment focus on Quality of Service(Qo S)and deployment costs,as its applications become more widespread,reliability and energy consumption cannot be ignored too.In this paper,a deployment model with four objectives of Qo S,deployment cost,reliability and energy consumption is constructed by considering multiple deployment models,and the problem is solved by using the many-objective large-scale evolution algorithm.In addition,the number of decision variables will increase with the increasing complexity of the optimization problem,it is necessary to improve the grouping algorithm to reduce the dimension of decision variables and design an improved algorithm to achieve better results.We will study an improve grouping algorithm to reduce the dimension of variable attributes,combine and improve different algorithm optimization methods,we proposed an improved many-objective evolution algorithm and apply it to solve the problem of 5G network deployment optimization.Finally,to improve the efficiency of the algorithm,this paper explores the parallel implementation of the improved algorithm.The main work of this paper includes:This paper propose a method of grouping decision variables based on coefficient of variation,combine and improve the optimization methods of different algorithms,and propose an improved parallel many-objective evolution algorithm combined with coevolution framework.We first used general test functions to verify the effectiveness of the improved algorithm,and then applied it to the problem of 5G network deployment optimization.The superiority of the algorithm is verified by experimentsSecondly,we study on the 5G network deployment model.An improved 5G network edge intelligent device deployment model is constructed,which considers four objectives: Qo S,deployment cost,reliability and energy consumption.Then,considering the impact of data peak,the improved algorithm is used for the first phase deployment of peak data,and then the dormancy strategy is formulated to make the nodes with low load go into dormancy to save resource consumption.Through experimental analysis,the improved algorithm can effectively reduce the value of each objective and compared with other advanced optimization algorithm,it has certain advantages,and the use of sleep strategy will not affect the quality of service of the system.Finally,to reduce the running time and improve the running efficiency of the algorithm,the parallel implementation of the algorithm is realized.