Research On Offloading Mechanism Of Vehicle Edge Computing Based On Deep Reinforcement Learning

With the rapid development of the Internet of vehicles,vehicles are turning from vehicles to intelligent terminals.The on-board network is facing the challenge of providing ubiquitous connections and high-quality services for many vehicles.The vehicle is equipped with on-board units with communication,sensing and computing capabilities.Vehicles can communicate with each other through wireless communication.However,the computing power of the vehicle is insufficient.Traditionally,cloud computing has strong computing power,but transmitting tasks to remote cloud usually leads to high delay and high energy consumption.In order to support delay sensitive and multimedia rich services in the Internet of vehicles,vehicle edge computing has become a promising computing technology to provide better computing services,which is computing and processing will be pushed to nearby edge nodes to assist processing tasks,such as vehicles,RSUs and BSs.Therefore,based on the expansion of computing resources,this paper studies the offloading mechanism of vehicle edge computing by taking system delay,system energy consumption and system utility as the indicators to measure the comprehensive performance of the system.Firstly,this paper introduces the vehicle network architecture and vehicle edge computing,which is the main model scenario factor in this paper.The purpose is to expand the computing resources of the vehicle network,save a lot of communication resources and reduce the transmission delay for the vehicle network.At the same time,the key technologies of vehicle networking communication are introduced,mainly including DSRC and C-V2 X.In order to solve the problem of resource management and task offloading decision of edge server in vehicle edge computing environment,the related concepts of deep learning algorithm and reinforcement learning algorithm are introduced to improve the comprehensive efficiency of the system.Secondly,this paper constructs a vehicle edge computing task offloading model based on DQN algorithm.In order to expand the computing resources of the Internet of vehicles,the concept of vehicle edge computing is introduced in this paper.In this model,the vehicle needs to process the tasks generated in each time slot.The processing methods are divided into local processing,offloading to adjacent vehicles,offloading to RSU and offloading to BS.In the offloading processing,the vehicle first needs to transmit the calculation task to the edge node through the wireless channel,then the edge node processes the calculation task,and finally transmits the calculation result back to the original vehicle.In order to reduce the system overhead,we introduce DQN algorithm to solve the problem of task offloading and processing,and make the optimal unloading decision to improve the quality of service.Finally,based on actor-critic algorithm,this paper proposes a vehicle edge computing task offloading model based on frame structure.Due to the limited computing resources of vehicles in terms of computing power,the possibility of completing tasks in time may be limited.The task can be offloaded to the roadside unit VEC server with stronger computing power,and the resource reservation server is used as the resource reservation of the VEC server to solve the above problems.In order to improve the convergence efficiency,we combine reinforcement learning with deep learning,and use actor-critic algorithm to transform the maintenance of value function table into the training of neural network model,so as to improve the training efficiency of model,improve the utility of VEC server and reduce the vehicle cost.There are 30 figures,2 tables and 73 references in this paper.