Research On Task Migration Strategy Of Mobile Edge Computing For Internet Of Vehicles

In the Internet of vehicles,intelligent connected vehicles can obtain more accurate and comprehensive traffic information with the help of its equipped computing unit and communication technology,and realize intelligent vehicle control,traffic management,safety warning and interactive applications.With the upgrade of intelligent connected vehicles,emerging on-board applications have higher requirements for computing power,and the computing power of on-board units is far from enough.By deploying the MEC server on the edge of the network and utilizing its rich computing resources,mobile edge computing technology can provide computing and unloading services for vehicles,which is an effective way to overcome the problem of insufficient computing power and limited communication resources of vehicles.In order to solve the problem of task unloading in edge computing of internet of vehicles,a Multi-objective Task Unloading Optimization Algorithm based on Load Prediction is proposed to reduce task delay and achieve load balancing of edge servers.First,the load of MEC server is predicted by Load Prediction based on AGA-LSTM algorithm,and the load change of MEC server is sensed in advance to solve the problem of unloading lag;Then,taking the delay and MEC server load balancing as the goal,and comprehensively considering the communication environment,computing resources and task volume,a multi-objective optimization model is constructed;Finally,the optimal task unloading strategy is obtained by Non-dominated Sorting Genetic Algorithm-Ⅲalgorithm.The simulation results show that the Load Prediction based on AGA-LSTM algorithm can predict the load of MEC servers more accurately.The Multi-objective Task Unloading Optimization Algorithm based on Load Prediction reduces the task delay by 1.7%,7.3%,12.4%,and 17.5% compared with the Multi-objective Task Unloading Optimization Algorithm,NSGA2-based task offloading strategy,Approach to Qo S-based Task Distribution,and All offloading strategy,and achieves significant advantages in MEC server load balancing to solve the problem of unbalanced MEC server load.In addition,by comparing the average transmission time delay and unloading rate of each communication cell horizontally,Multi-objective Task Unloading Optimization Algorithm based on Load Prediction can develop differentiated unloading schemes according to the communication environment and the number of vehicles in different communication cells.Due to the mobility of the vehicle and the limited coverage of the edge computing server,the offloading of the service to the MEC server can be completed before the vehicle has moved out of the MEC server range and established a new communication connection with another MEC server.In this case,it is necessary to determine whether the service needs to be migrated according to the location of vehicles to ensure the quality of service.To solve this problem,this paper aims at the migration requirements of vehicle services in the Internet of vehicles scenario,fully considers the mobility characteristics of vehicles combined with the road network model,and develops service migration strategies with the goal of minimizing delay and MEC server load balancing.A Markov Decision Process(MDP)model based on discrete variables is proposed to solve the service migration problem.Compared with other service migration algorithms,it is proved that MDP service migration algorithm has better performance results in terms of delay,load balancing and service migration times.Compared with Multi-Attribute Decision making strategy,MDP strategy has lower load balancing standard deviation and service migration times by 4.8% and 7.2%,respectively.In order to further test the performance of the algorithm,this paper built an experimental platform around the task unloading of Internet of vehicles,based on on-board equipment,roadside equipment,display and control terminals,and realized V2 X application development,communication transmission and other functions.Based on the experimental platform,the communication performance of the network of vehicles under different communication distances and vehicle speeds was tested and compared.Then,the communication delay was tested in the real campus road network environment.The optimal unloading strategy was obtained by inputting the multi-objective optimization unloading strategy algorithm based on load prediction,and the feasibility of the algorithm was further tested.