Research On Computing Offloading And Resource Allocation In Internet Of Vehicles Based On MEC

With the development of the Internet of Vehicles technology in 5G,vehicles perform real-time environment perception tasks to obtain accurate road information and make behavioral decisions during driving.The use of multi-access edge computing technology can alleviate the contradiction between unlimited computing tasks and limited computing resources to a certain extent.However,compared with ordinary mobile users,vehicle terminals have the characteristics of fast driving speed and variable driving environment,which makes task offloading more complicated,and puts forward higher requirements for low latency and high reliability of vehicle terminal resource management.On the one hand,using the reinforcement learning algorithm to train the unloading strategy and optimize the resource allocation method can realize the efficient sharing of vehicle network computing resources and meet the intensive computing requirements;A more efficient data processing method can shorten the computing delay of the edge computing network and improve the overall service level of the system.Based on the above research,a joint computing offloading strategy is proposed according to different task types processed by vehicle terminals and roadside units,which is respectively optimized for the V2I and V2V communication methods in the Internet of Vehicles.The specific content is as follows:First of all,in view of the large amount of original image data collected by the vehicle,it is easy to cause network overload and affect the transmission delay and collaborative perception timeliness.A joint task offloading method is proposed to relieve the pressure on the wireless communication network while reducing the loss of hardware equipment;in the transport layer,the optimal offloading strategy is obtained by using the deep Q network algorithm with the experience replay and the target network.Complete the task with low computing delay and sends the calculation result back to the vehicle terminal to optimize the V2I communication in the vehicle network;at the application layer,the improved regularized backtracking matching tracking algorithm is used to reconstruct the original information,and through the Simulation experiments on real datasets demonstrate the effectiveness of the proposed algorithm.Secondly,aiming at the problem that intelligent vehicles of different types and capacities and unpredictable vehicle topology make the agent unaware of part of the environment,the focus is on research and analysis of the collaborative relationship between different vehicles in the vehicle edge computing network.First,based on the gravity model improved quality-resource aggregation model is proposed,and vehicles are aggregated by measuring the relationship between vehicle computing tasks and computing resources.Then,a multi-agent deterministic gradient descent reinforcement learning algorithm based on distributed scheduling is used to optimize V2V Computational task offloading and edge resource allocation for communications minimize offloading costs and efficiently integrate service resources.