Research On Resource Scheduling And Task Guarantee Mechanism For Intelligent And Connected Vehicle Environment

With the promotion of smart city,Internet of Things,artificial intelligence and other technologies and the guidance of national policy to build a powerful country in transportation,Intelligent and Connected Vehicle(ICV),which integrates intelligence and network connection,has become a cutting-edge direction for the global automobile industry reform.In the ICV environment,for the massive,real-time and heterogeneous task demands bestowed by intelligent and connected functions on vehicles,it is of great significance to realize efficient resource scheduling before task execution and reliable service guarantee process during task execution,thus improving the dynamic network performance.However,when scheduling service resources to tasks,it is difficult for vehicles with limited space and cost to deploy a large number of resources locally to provide services,and remote cloud resources are easily suffered with the problems of prolonged data transmission and low timeliness serving.At the same time,the difficulties such as connection changing,service interruption and long-distance content reply caused by vehicle mobility also pose many challenges to the task service guarantee process.To overcome the above problems,this paper conducts an in-depth research on the resource scheduling and task guarantee mechanism for ICV environment.Firstly,combing Software-defined Networks(SDN)with Mobile Edge Computing/Multi-access Edge Computing(MEC)technology,a foundational ICV environment model covering edge layer,SDN control layer and cloud layer is designed.The edge layer realizes the basic network connections through the network links among multiple nodes such as vehicles,pedestrians and Road Infrastructures(RIs);while the SDN control layer performs SDN cell management and completes some intelligent resource collecting and scheduling works upon its efficient and flexible data integration ability;the cloud layer is responsible for meeting the complex connection and intelligent requirements of the whole vehicular networks.Next,following the task lifecycle,before the execution of ICV tasks,two resource scheduling algorithms in different scenes are proposed;during the execution of ICV tasks,for the interrupted tasks caused by vehicle mobility,a task guarantee mechanism based on dynamic service migration is designed,and for the content request tasks,a task guarantee mechanism based on pre-caching is designed.The specific research contents are as follows.1)In vehicle sparse scenes,aiming at the problems of unbalanced resource utilization and task execution efficiency in the edge server resource scheduling process,an edge server resource scheduling algorithm based on future availability assessment is proposed,which aims to minimize the average task serving time in SDN cells and also considers the future edge server resource availability and network performance parameters.At first,a neural network model is used to predict the future demands and further assess the availability of edge server resources;then,on the basis of optimizing task serving time,the impact of task serving process on the network performance is also considered to form a comprehensive objective function with soft and hard constraints;finally,an edge server resource scheduling algorithm based on future availability assessment is designed for the objective function and its constraints.Simulation results show that the above algorithm can effectively improve the efficiency of vehicle task execution and balance the edge server resources allocation.2)In vehicle intensive scenes,motivated by the demand of task cooperation among vehicles and the problem that vehicle mobility advantage is easily ignored,a vehicle service resource scheduling algorithm based on spatio-temporal trajectory prediction is proposed,which explores how to give full play to the advantages and opportunities brought by the vehicle motion correlations to the task execution process in the same RI serving area.Firstly,a vehicle trajectory prediction model is designed from time and space two dimensions;next,the position and trajectory deviation measurements are defined to obtain the corresponding trajectory coincidence degrees among vehicles in a period of time;finally,combined with the objectives and constraints,an optimization function is established,and a vehicle service resource scheduling algorithm is further proposed under the guidance of the optimization function.The simulation results show that the above trajectory prediction model achieves the highest accuracy,moreover,the designed vehicle service resource scheduling algorithm has the optimal task serving efficiency and success ratio under different vehicle density and motion modes.3)For the task interruption problem caused by vehicle mobility,a task guarantee mechanism based on dynamic service migration is designed,which studies on the process of following the dynamic vehicle positions to migrate tasks among SDN cells during the task execution process,thus ensuring the stability and continuity of dynamic services.Firstly,based on a general MDP model,the existence of state threshold triggering migration action in MDP-based service migration scheme is proved;secondly,the functions used in the MDP model,especially the revenue function,are defined in detail;next,the value iteration process of Bellman equation is applied to calculate the task migration revenues in each state and solve the migration threshold to guide the action;at the same time,the parameter updating process in vehicle network environment is also designed.Simulation results show that the proposed dynamic service migration method can effectively reduce the task response time,data packet loss ratio and the total number of service migration in the vehicular networks.4)For the problem of insufficient service stability and reliability of vehicle content request tasks,a task guarantee mechanism based on precaching is designed,which further studies the content pre-caching problem by combing with the residence time of vehicles in SDN cells,so as to ensure the reliability of the dynamic content services.Firstly,the cell congestion degree is defined upon the vehicle historical trajectories,speeds and traffic state division standards,and the future vehicle speeds and trajectories are also predicted;then,combined with the comparisons of the estimated cell residence time and content download delay,the pre-caching cells can be selected;finally,using the centrality,load degree of each edge server and content popularity in the selected cell,an optimal edge server will be selected to cache the content.Simulation results show that the proposed pre-caching method can reduce the serving time of content requests in vehicular networks and improve the cache hit ratio.