Research On Task Offloading Method Of Edge Computing In Internet Of Vehicles Based On Reinforcement Learning Strategy

With the rapid development of society,the number of vehicles is rising rapidly,which also leads to the aggravation of traffic jams and frequent traffic accidents.Intelligent transportation--Internet of vehicles should also be developed.However,due to the particularity of the task of Internet of vehicles,its characteristics of high bandwidth,low delay and high reliability,the traditional cloud computing model has been unable to meet its needs.As an extension and redevelopment of mobile cloud computing,mobile edge computing has a good correlation with the Internet of vehicles.As an important research direction in edge computing,computing offloading also poses great challenges.Based on reinforcement learning and game theory,two new offloading strategies are studied and proposed in this paper.The simulation results show that the new offloading strategy is better than the traditional one in terms of delay,energy consumption and total cost.Two new strategies are proposed in this paper:1.Aiming at the task unloading system of Internet of vehicles,considering the situation of multiple MEC servers in modeling,a deep reinforcement learning unloading scheme is proposed Strategy,DRLOS),which improves the traditional Q-learning algorithm,combines deep learning with reinforcement learning to avoid the dimension disaster in Q-learning algorithm and achieve better results.Simulation results show that the proposed algorithm has better performance on delay,energy consumption and total system overhead under different number of tasks and wireless channel bandwidth.2.The problem of multi-user task offloading in dynamic environment is studied.Due to the limited communication and computing capacity of the edge server,when multiple users unload tasks to the same edge server,resource competition will occur.Therefore,we regard this problem as a game model.Using replicator dynamics to analyze the process of user selection strategy,it is proved that there is a unique Nash equilibrium in the multi-user computing offload model.In the actual application scenario,a game theory strategy based on reinforcement learning(GTSRL)is designed.Finally,experiments are used to verify the convergence and performance of the proposed algorithm in multi-user scenarios.