Research On Task Migration Strategy In Mobile Edge Computing

With the development of artificial intelligence,5G/6G and other emerging technologies,automatic driving,online games,augmented reality and other applications with intensive computing and memory access and low latency are increasingly in demand at the end devices.However,devices with limited resources are difficult to meet the demand of computing resources for the above applications.However,the core network of transmission between cloud data center with central architecture and devices is congested,which is difficult to support the real-time demand.To this end,mobile edge computing,which performs computing near the network edge of user,has been proposed by researchers.At present,extensive works have been done for edge computing,however,most of them focus on task offloading strategy with quasi-static models,ignoring the mobility characteristics of users.Whether to migrate tasks dynamically based on the user's movement trajectory can not be ignored in further improving Qo S.This paper mainly focuses on the task migration in edge computing,and the specific research contents are as follows:(1)To solve the problem of task migration in edge computing,a dynamic edge computing scenario with multiple edge nodes and multiple mobile users is constructed,in which the mobile users dynamically move among multiple edge nodes after offloading their tasks to the edge nodes,and the tasks of the mobile users have a strict deadline for completion need to be guaranteed.Based on this scenario,the optimization objective is to maximize the number of tasks completed within the deadline.Through the mathematical modeling of the scene,it is proved that the maximum optimization problem is NP hard.In order to solve the optimization problem,the mobility information of users is exploited to analyze the three situations of the original optimization problem,and the migration threshold is defined.The solutions under the three situations and theoretical analysis are given.Based on those,a group migration algorithm,GM,is proposed,which maximizes the number of tasks completed within the deadline.The simulation results verify the good performance of the group migration algorithm,which can achieve a performance improvement of35%-75% compared with other benchmark algorithms.(2)To solve the problem of task migration under the constraints of migration energy consumption in edge computing,a dynamic edge computing scenario with multiple edge nodes and multiple mobile users is constructed,in which the user's service quality and migration energy consumption of edge node are considered,the optimization objective is to minimize the average completion time of all tasks under the constraints of migration energy consumption.To solve the optimization problem,the task migration problem is modeled as a Markov decision process based on the assumption that the mobility of users is Markov.A task migration algorithm based on multi-agent reinforcement learning was proposed,to minimize the average completion time of all tasks under the constraints of migration energy consumption,and the convergence of the algorithm was analyzed.A large number of simulation experiments were carried out to evaluate the performance of the algorithm.Compared with other algorithms,the experimental results show that the task migration algorithm based on multi-agent reinforcement learning can reduce the average completion time of the task by 30%-50%.