Research On Prediction Task Offloading And Serving Handover Strategy Of Vehicular Edge Computing Networks Based On DRL

With the rapid development of 5G mobile communication and intelligent transportation technology,many emerging vehicular applications（such as autonomous driving,real-time navigation,intelligent identification,etc.）have emerged in recent years,which require a lot of computing and storage resources.However,in the highly dynamic vehicular communication environment,the resources of vehicles are too limited to complete these urgent tasks under the requirements of low latency and high reliability.At present,MEC（Mobile Edge Computing）has become a promising task computing architecture.VEC（Edge vehicular computing）network provides services for vehicle tasks by deploying RSUs（Road Side Units）or requesting moving vehicles on the road as task processing servers.By sinking cloud computing services to the edge,the task transmission distance is shortened,and the task execution delay is greatly reduced.It can not only meet the emerging 5G communication technology and the real-time task processing requirements of autonomous driving,but also meet the communication reliability of high-speed vehicles.However,the topology of the vehicle network in real-world roads is extremely unstable,and the available computing resources of roadside units and moving vehicles are also constantly changing.Whether the offloading task can be completed within the delay constraints and vehicle communication range is still unknown.Therefore,it is difficult to reasonably select the offloading target and processing method of the task in Io V（Internet of Vehicles）.This thesis mainly focuses on the task offloading and serving handover strategies of the Internet of Vehicles in different scenarios.The main work is as follows:（1）In the highway scenario,this thesis proposes a task offloading and serving handover scheme for VEC.On the highway,the computing resources and storage resources are relatively limited,and the vehicle travels at a fast speed.Therefore,the offloading tasks needs to be completed within the communication range.Firstly,according to the possible task offloading strategies,an offloading model is constructed and an objective function to minimize the total energy consumption is proposed.Then,the optimal problem is decomposed into two policy problems: task offloading decision problem and serving handover decision problem.Finally,an iterative optimization method is used to optimize the solution to the proposed problem.Simulations show that the proposed task offloading and serving handover scheme significantly reduces the total energy consumption of task processing.（2）For urban road scenarios,this thesis proposes an efficient task offloading scheme for VEC based on trajectory prediction,and focuses on the serving handover between adjacent RSUs.The moving vehicles can handle tasks locally,or they can choose to offload tasks to surrounding RSUs or cooperative vehicles for task processing.To reduce the latency of task transmission between vehicles and improve the stability of data transmission,a cooperative vehicle selection method based on trajectory prediction is proposed in this thesis.Then,we propose an efficient task offloading scheme based on deep reinforcement learning（DRL）,while the dynamically available computing and communication resources are considered jointly.The simulation results show that the proposed task offloading scheme has great advantages in improving the utility of vehicles.The research of this thesis can effectively improve the utility of vehicle task offloading and reduce the energy consumption and time of task processing.It can also help vehicles to select the task offloading strategies and serving handover strategies in the highly dynamic vehicular communication environment,improving the service experience of vehicle users.