Research On Application Technology And Algorithms Of Cooperative Edge Computing And Internet Of Vehicles

The intelligent connected vehicles and roadside infrastructures have promoted the development of the Internet of Vehicles(Io V)which plays a pivotal role in helping to build a safe,efficient,and environmentally friendly transportation environment.Although the vehicles’ local computation and storage capabilities have been significantly improved,it is still difficult to handle intensive Io V tasks within the specified delay.Therefore,some tasks with high computational complexity and energy consumption are offloaded to the cloud computing platform for processing and then returned to the vehicles.Although cloud computing effectively relieves the pressure on vehicles,it also greatly increases the latency overhead,core network load,and the risk of privacy leakage due to long-distance transmission and centralized processing.To compensate for these shortcomings in cloud computing,an edge computing approach closer to the vehicles shows corresponding advantages.Edge computing provides vehicles with elastic computation,communication,and storage resources.Compared with cloud computing,edge computing also effectively reduces latency,alleviates data congestion in the network,and protects vehicle privacy to a certain extent.With the exploration of the application of cooperative multi-access edge computing(MEC)and Io V,MEC as a new computing paradigm in the Io V has gradually presented some new problems to be solved.For example,the architecture design of cooperative MEC and Io V system,the problem of Io V task offloading and computation resource allocation,the problem of vehicle privacy preservation in the process of offloading,and the overall performance evaluation test of the system.In the paper,research is conducted to address the above four challenging problems.The main research content and innovation are as follows.(1)Research on the architecture of cooperative MEC and Io V system.The current standardized reference architecture for MEC cannot meet the specific needs of the vertical industry as Io V,and the standardized system architecture of cooperative MEC and Io V has not been formed.This part proposes a flexible system architecture that can meet the end-to-end process connectivity requirements of Io V applications.The architecture includes the four-level main participation elements involved in the Io V applications and the four interfaces required by accessing the MEC platform for each element.According to the interaction relationship between elements and the defined interfaces,this research further sorts out the common end-to-end process of cooperative MEC and Io V applications and summarizes the application of multiple types of scenarios to the architecture.Finally,the system architecture is analyzed in terms of functional and performance advantages in many aspects.(2)Research on joint optimization of task offloading and computation resource allocation in the cooperative MEC and Io V system.To address the problem of system delays and high computation resource costs due to unreasonable task offloading decisions in a multi-vehicle competitive environment,this part proposes a reasonable offloading and computation resource allocation scheme for multiple Io V tasks under the condition of limited computation resource,to effectively optimize the end-to-end delay and computation resource overhead in the system.In order to meet the low latency requirements of Io V applications,this part deploys a MEC server that can effectively improve the data processing capability of the vehicles in the system model,allowing the vehicles to perform binary offloading decisions under the condition of load balancing of the MEC server.Considering the high-speed mobility of vehicles,the randomness of the Io V task generation,and the highly dynamic change characteristics of resources in the system model,when solving the mixed integer nonlinear joint optimization problem,this part firstly models the optimization problem as a Markov Decision Process(MDP)and then proposes a multi-agent Deep Reinforcement Learning(DRL)algorithm to obtain the offloading and computation resource allocation optimization strategies under high-dimensional continuous state space.Experimental results show that the proposed method can not only effectively meet the computing requirements of the Io V tasks,but also significantly improve the comprehensive performance of the system in terms of delay and computation resource overhead.(3)Research on collaborative optimization of task offloading and computation resource allocation considering vehicle privacy preservation.In order to reduce the risk of vehicle privacy leakage caused by data sharing during the Io V task offloading,this part proposes an Io V task offloading and computation resource allocation strategy that integrates the privacy preservation mechanism,to protect the privacy of the vehicle to the greatest extent and make reasonable use of the computation resources in the system.The system model builds a two-layer heterogeneous vehicular edge computing network including a security authentication center,in which vehicles can protect identity and location privacy through pseudonym changing,while allowing vehicles to offload tasks to idle neighbor vehicle nodes to improve the utilization of computation resources in the system,and also add partial offloading mode to reduce the task execution time.Given that Io V tasks have different priorities,tasks with higher priorities are generally more sensitive to latency.When designing the optimization problem,the influence of factors such as communication silence during privacy preservation,task priority,and vehicle motion state on the optimization scheme is comprehensively considered.To avoid dimensional disasters caused by complex state space,a decentralized multi-agent DRL collaborative optimization algorithm for safe offloading and computation resource allocation is proposed.The simulation results show that the proposed scheme has good convergence,and has more satisfied safety performance,service quality,and high computation resource utilization compared with the other four algorithms.(4)System-level test research of cooperative MEC and Io V systems.In order to accelerate the large-scale promotion of cooperative MEC and Io V applications,it is necessary to carry out confirmatory tests of existing system-level research schemes as soon as possible.To address the problem that the current device-level test indexes and simulation test methods in the laboratory cannot meet the application test requirements of cooperative MEC and Io V systems in real scenarios,this part proposes a test scheme for the practical application requirements of cooperative MEC and Io V cascade systems.The proposed test scheme mainly includes four categories of test indexes and test methods corresponding to indexes.Firstly,the definition and selection basis of each index is given.Secondly,quantifiable test methods are designed based on the two ground truth test systems built by the China Academy of Information and Communications Technology.Finally,field test experiments are carried out on various system-level research schemes.The test results show that the proposed test scheme can quantitatively evaluate the performance of different research schemes under each test index,which meets the need to test and evaluate existing system-level solutions in real scenarios.