| With the rapid development of 5G network and computer vision applications,a large amount of video data in the Internet of Vehicles is used for content analysis to boost drive safely.On the one hand,the tasks based on video content understanding are usually accompanied by huge data volume and huge computing power requirements.Vehicles resolve the conflict with their limited capabilities by offloading such computing-intensive applications to Mobile Edge Computing(MEC)servers for video content understanding.On the other hand,existing resource allocation schemes based on Quality of Service(QoS)or Quality of Experience(QoE)may not be the best choice for the purpose of video content understanding.Therefore,it is necessary to consider communication and computing resources to design a new efficient joint resource allocation scheme.The topic of the thesis comes from Beijing Natural Science Foundation project "Research on resource allocation algorithms of Internet of vehicles based on video content understanding driven by dynamic spatio-temporal data(4202049)".The research work focuses on the joint allocation of communication resource and computing resource in the Internet of Vehicles,analyzes the differences between the joint resource allocation based on video content understanding services and the traditional resource allocation,and proposes the QoC-based joint resource allocation algorithm in the Internet of Vehicles,so as to support the demand of high reliability and low delay for video content understanding services.The main research and achievements are as follows:(1)This thesis comprehensively describes the research status of Mobile Edge Computing technology and resource allocation in the Internet of Vehicles.Firstly,the necessity of introducing MEC into the Internet of Vehicles is explained,and then the basic concept and application scenarios of MEC are introduced.Then it introduces the research status of resource allocation in the Internet of Vehicles in detail,analyzes and points out the existing problems and challenges in the research of resource allocation,and then gives the research direction of this thesis.Finally,the research status of the application of machine learning algorithms in resource allocation is introduced,which lays a foundation for the research of QoC-based joint allocation algorithm of communication and computing resources in the Internet of Vehicles.(2)In view of the huge amount of video transmission and video content understanding services in the Internet of Vehicles,which brings great pressure to the traditional wireless communication resource and MEC servers,and the existing resource allocation schemes based on human perception can not effectively improve the accuracy of video content understanding,this thesis proposes a QoC-based joint resource allocation algorithm in the Edge-dominated computing scenario.Firstly,the model of target detection accuracy under the constraints of spectrum and computing resource is constructed,and then the unified expression of video content understanding accuracy under communication and computing resource constraints is derived.Then,considering the real-time nature of resource allocation and the variability of environment,a Multi-agent Distributed Q-Learning algorithm is proposed to solve such multi-constraint nonlinear programming problem.Finally,the results of QoC-based joint resource allocation are analyzed by MATLAB simulation,and the performance of the proposed QoC-based joint resource allocation algorithm is evaluated by comparison.(3)Aiming at the problem that the joint resource allocation based on deep learning model has different requirements and utilization for communication and computing resources,in order to further improve the accuracy of video content understanding and effectively reduce the system delay,a QoC-based joint resource allocation algorithm in the Vehicle-edge synergy scenario is proposed.Firstly,it analyzes the differences between the joint resource allocation based on deep learning model segmentation and the traditional MEC research on the requirements and utilization of communication and computing resources.Then,jointly considering the Deep Neural Network segmentation,communication and computing resources allocation,this thesis proposes a joint resource allocation algorithm based on Distributed Q-Learning in the Vehicle-edge synergy scenario.Finally,through the MATLAB simulation,the resource allocation algorithm is compared with the other two scenarios,and then the performance and system delay of the resource allocation algorithm in the three scenarios are evaluated.This thesis proposes QoC-based a joint resource allocation algorithm in the Internet of Vehicles.In the case of limited communication and computing resources,it can maximize the performance of video content understanding and effectively reduce the delay,so as to achieve the demand of high reliability and low delay.It provides a new idea for the new joint resource allocation scheme of computing intensive services in the Internet of Vehicles. |
