Research On Offloading Strategy Of Mobile Edge Computing For Internet Of Vehicles

In Internet of vehicles,the resource-constrained vehicle terminal equipment cannot meet the requirements of new vehicle applications for resources and time delay.For this reason,the Mobile Edge Computing(MEC)technology is introduced into Internet of vehicles.MEC deploys and provides application service environments and cloud computing capabilities at the edge of the network to meet the expansion needs of vehicle terminal equipment computing capabilities and shorten the processing delay of vehicle applications.However,the limited computing power of MEC,the cost of the MEC offloading system,and the complexity of Internet of vehicles environment seriously affect the performance of offloading tasks to the edge of the network.This thesis studies the offloading strategy of MEC for Internet of vehicles,focusing on the problem of computing task offloading in them.The main work of the thesis has the following two aspects:1.With considering the existing problems that the demands of processing a large number of new vehicle applications are difficult to meet in a short time by resourceconstrained vehicle terminal equipment and the dynamic characteristics of the offloading nodes cause the uncertainty of the candidate offloading nodes in the complex and changeable environment of the Internet of vehicles,this thesis proposes an edge computing offloading strategy based on reinforcement learning for Internet of vehicles.Firstly,the strategy combines the device link time,communication radius and other factors to develop the offloading node discovery mechanism.Secondly,the utility function is established by combining task processing delay and cost,in addition,the problem of task offloading and computing resource allocation is transformed into the problem of optimizing system utility.Finally,an intelligent node selection offloading algorithm is proposed on the basis of the offloading node discovery mechanism and reinforcement learning method to achieve intelligent offloading of tasks and reasonable allocation of computing resources.The experimental results show that the proposed offloading strategy can achieve high system utility in the urban street scene of Internet of vehicles.2.In the scene of one-way through road with sidewalk,there are some problems,such as vehicle mobility leads to the discontinuous transmission of task offloading,besides,the complex and changeable environment of the Internet of vehicles and the datadependent offloading tasks affect the offloading performance,based on which the edge computing offloading strategy is proposed in this thesis to keep the offloading continuity in the Internet of vehicles.Firstly,the breadth-first search(BFS)algorithm and experience are adopted to determine the amount of data in the task unit according to the task type and its related module data dependency graph.Secondly,the task is divided into several task units with the same amount of data,and the number of task units pre-allocated to each cell is determined according to the environmental conditions and restrictions of each cell.Finally,the computing task is offloaded in order to minimize the total cost of completing the computing task.Experimental results show that the proposed edge computing offloading strategy has excellent performance in cost efficiency.