Research On Offloading Strategy Of Low Latency Tasks In Mobile Edge Computing

With the rapid development of the Internet of things and communication technology,the vehicle network is facing the challenge of higher speed,more reliability and lower delay.Augmented reality / virtual reality,automatic driving,collision warning and other emerging intelligent application services with computing intensive and highly delay sensitive make the challenge more urgent,and the limited computing and storage resources of the vehicle terminal can not meet the increasing demand Growing demand.In addition,the available spectrum resources are not enough to support the wireless services of a large number of vehicle users.For this reason,Mobile Edge Computing(MEC)has become a new network architecture to improve user experience,and has been widely concerned.MEC technology sinks the business to the edge of the network,improves the nearby it service and cloud computing capacity for vehicle users,greatly reduces the response delay,and meets the low delay requirements of vehicle users with limit ed resources.The tasks that are computationally intensive or delay sensitive on the vehicle equipment are unloaded to the nearby MEC server through the wi reless network,and the computing tasks are completed by relying on the abundant resources on the clo ud server,so as to reduce the delay of task response and solve the problems such as the computing capacity and storage capacity of the vehicle terminal.T herefore,in the vehicle intensive environment,task unloading strategy is the key to improve the sys tem performance,and task unloading strategy is very important.In this paper,the architecture of vehicle network based on MEC is designed,and the integr ation of software defined network(SDN)and mobile edge computing is introduced,which can provide more flexible centralized control and resource management for vehicle network.The main research contents of this paper are as follows:1.The problem of unloading strategy and task scheduling between single channel vehicle users is studied.Firstly,the vehi cle network is modeled to optimize the service delay of MEC system with multiple independent tasks.The unloading strategy is not only to decide whether the tasks of vehicle users are unloaded,but also to schedule the unloading tasks to get the execution order,so as to achieve the optimal system delay.In order to reduce the time complexity,this paper proposes a dual flow waterline scheduling algorithm to schedule the unloading task,and proposes an improved discrete binary particle swarm optimization al gorithm combined with genetic algorithm to optimize the unloading strategy.The simulation results show that the proposed solution can effectively reduce the task response delay and improve the system performance.2.The problem of unloading strategy and s pectrum selection among multi-channel vehicle users is studied.Firstly,this paper studies the task unloading of MEC system in the multi-channel wireless interference environment.The Road Side Units(RSUs)has the cognitive function and can sense the avai lable spectrum of the surrounding environment.The vehicle users not only choose whether to unload,but also choose the transmission channel.Firstly,a hi erarchical algorithm based on K-means is proposed to cluster the vehicles.Different communication fr equency bands are used in the cluster to reduce the interference among the members of the cluster.Finally,due to the competition among vehicle users,each vehicle user is regarded as the player,and the unloading strategy of vehicle users associated with different roadside units is constructed as an n-person game model.A game strategy optimization algorithm based on correlation equilibrium is proposed,an d the relevant equilibrium solution of the game is obtained through regret matching mechanism.Simula tion results show that the proposed scheme can achieve fast convergence of correlation equalization and reduce system delay.