Research On Task Offloading And Handover Algorithm In Mobile Edge Computing

With the development of technology,the demand for compute-intensive applications in mobile devices is increasing.Executing such a compute-intensive application on a mobile device can result in slower response times,which reduces the quality of service for users.Mobile Edge Computing(MEC)has become an effective way to solve this problem.MEC is done by deploying the server on the edge of the network that is far away from the cloud server.Compared with mobile devices,it has richer computing and storage resources,and can provide data processing services for different types of users,thereby improving the quality of service of users.Due to the small coverage of the MEC server,the total revenue it receives is greatly affected by the mobility of the user’s device.Therefore,in the case of limited server resources.How to design an effective task offload strategy based on the mobility of the user device to maximize the total benefit of the system is of great significance,and it is also the main research content of this thesis.The highly dynamic topology of the vehicle communication system and the time-varying available computing resources challenge the efficient task offloading of the vehicle.Recently,a number of task offloading methods based on Deep Reinforcement Learning(DRL)have been proposed.However,these methods are less adaptable to the new environment and cannot adapt well to high-speed moving vehicles.In this thesis,a MEC adaptive task offloading method based on trajectory prediction is proposed,which only requires a small amount of gradient update and samples to quickly adapt to the new environment.This method can better adapt to the environmental time variability brought about by the rapid movement of the vehicle.This method selects candidate vehicles that can be offloaded by trajectory prediction method,and converts the computational offloading process into a sequence prediction process,which effectively represents the strategy by sequence-to-sequence neural network.Experimental results show that the proposed method has great advantages in terms of the utility of task offloading.In the real world,most users use the application in a mobile state,so when the user moves from the service scope of one edge server to the service scope of another edge server,in order to ensure the quality of service and the continuity of the connection,it is necessary to perform a computing handover.Many studies have proposed host-based fast handover schemes,host-based solutions have the problems of long handover delay and large amount of packet loss,and it is difficult to be widely used.This thesis proposes a zoning minimization handover method based on software defined network(SDN)network control,the advantage of which is that mobile devices do not participate in the handover process,and the metropolitan area network is divided into multiple disjoint clusters to minimize the number of possible handover between different MEC regions,thereby reducing the total handover delay.Finally,experimental results show that the proposed method is superior to the existing scheme in terms of handover delay and packet loss,and significantly reduces the total number of handover and improves the average handover delay.