Research On Vehicle-road-cloud Cooperati-on Oriented Joint Optimization Of Proximity Detection Latency And Energy Consumption

The development of automatic driving and assisted driving has given birth to the problem of proximity detection in road networks,which plays an important role in ensuring safe driving.Proximity detection refers to whether two vehicles are close to each other in a short time.It is a timesensitive task.However,in actual scenarios,the battery power and computing ability of vehicles are limited.Therefore,how to solve this problem with low latency and low energy consumption is an important issue worth studying.Combining the advantages of vehicle-road-cloud collaboration,this thesis proposes solutions to jointly optimize the latency and energy consumption in the road network proximity detection scenario,so as to achieve the goal of real-time continuous detection(improving the real-time of proximity detection)and reducing the energy consumption of intelligent terminals.The work is as follows:(1)Research on joint optimization of proximity detection latency and energy consumption based on genetic algorithm.In the multi-user and multi-server scenario,a proximity detection computation offloading scheme based on mobile edge computing(MEC)is proposed,including local computing and offloading to the edge server computing.The computation offloading problem of joint optimization of latency and energy consumption in the proximity detection scenario is modeled as a constrained multi-objective optimization problem(CMOP)to minimize the average latency and energy consumption of the entire system.Considering the complexity of this problem,the second generation nondominated sorting genetic algorithm(NSGA-II)is used to solve the multiobjective optimization problem.The experimental results show that the NSGA-II algorithm is robust under the influence of different parameters,and can ultimately find Pareto optimal solution,which can meet the different needs of users.In addition,a pruning strategy based on time distance in dynamic road network is proposed,which can filter out users who do not need to be detected(that is,can judge whether they are close without detection).The experimental results show that the pruning strategy is effective and can significantly reduce the latency and energy consumption.(2)Research on joint optimization of proximity detection latency and energy consumption based on deep reinforcement learning.Considering that the advantages of AI in analysis,judgment and control are more suitable for the dynamic and time-varying complex environment of the Internet of Vehicles,AI technology and edge computing are organically incorporated into the Internet of Vehicles architecture.In order to solve the problem of the integration of data communication and intelligent decisionmaking,a communication-computation integrated proximity detection architecture for the Internet of Vehicles is proposed.In the multi-user and single-server scenario,a computation offloading scheme is proposed,including local computing,offloading to the edge server computing and offloading to the surrounding idle vehicles computing.The computation offloading problem of joint optimization latency and energy consumption in the proximity detection scenario is modeled as a constrained multiobjective optimization problem to minimize the latency and energy consumption of the target vehicle(i.e.,single-vehicle optimization).A simple and effective method DDPG-CMOA is proposed to solve this problem,including DDPG-based proximity detection algorithm(PDBDDPG)and dynamic road network update algorithm(DRNU).At the same time,multi-objective evolutionary algorithms(MOEAs)are also used to compare the performance with DDPG-CMOA.The experimental results show that the DDPG-CMOA method has short running time,strong generalization ability and good convergence and diversity.(3)A Django-based visulization system for joint optimization of proximity detection latency and energy consumption is developed and implemented.According to the user’s choice,it can provide the user with appropriate offloading decisions and proximity detection results,reduce the task execution latency and energy consumption,and improve the user experience.The Django framework technology is used as the Web application framework.Highcharts realizes interactive chart display,realizes the user recommendation process based on the deep reinforcement learning algorithm DDPG-CMOA,and applies the front-end technology to display the system interface.In summary,this thesis proposes two solutions for jointly optimizing latency and energy consumption in road network proximity detection scenarios,and develops a visualization system based on Django,achieving the goal of real-time continuous detection and reducing energy consumption of intelligent terminals.