Research On Task Offloading And Resource Allocation Strategy In MEC Networks

With the development of communication technology and the application of artificial intelligence industries,new network scenarios such as the internet of vehicles and unmanned aerial vehicle(UAV)networks are rapidly emerging.Devices such as smart vehicles and UAVs,have the ability of data sensing and processing,which benefit the deployment of control and analysis applications.Meanwhile,with the explosive growth of datas,it is difficult to meet the real-time requirements for data processing,due to the limited computation capability at end devices.In this case,mobile edge computing technology has been widely introduced.Tasks can be offloaded to the edge server for processing to cope with the problem of insufficient computing power at end devices,and the delay for data processing can be greatly reduced.However,the traditional network mainly focuses on communication,whose core function is the transmission of collected datas,which lacks the consideration on efficient use of ubiquitous computing resources at network edge.Therefore,how to give full play to the synergy of communication and computing to achieve efficient processing of sensing datas,is still a problem worthy of study.This thesis studies the task offloading and computing resource allocation strategies in mobile edge computing networks.First,aiming at the change of network topology in vehicular edge computing system,considering the differences of communication mode and computing cost between vehicles and roadside unit,the mobility-aware cooperative task offloading and resource allocation strategy is studied.Furthermore,aiming at the asynchronous arrival of multi task flow in UAV assisted real-time monitoring applications,considering the system's demand for information timeliness,the asynchronous computation offloading strategy based on deep reinforcement learning is studied.The main contributions of the thesis are as follows:(1)For on-vehicle computing-intensive applications,with the goal of optimizing task completion delay and cost for communication and computation in vehicular edge computing network,a matching-based collaborative task offloading and computing resource allocation scheme is designed,taking into account the unreliability of task offloading and result feedback caused by high mobility of vehicles.Simulation result shows that the proposed scheme significantly reduces the task delay and cost compared with other strategies,and greatly improves the reliability for task offloading and result feedback.(2)For UAV assisted real-time monitoring applications,the goal is to optimize the timeliness of information.Taking into account the task flow generated continuously at UAV and the asynchrony of task arrival between UAVs,a scheme of asynchronous task offloading and computing resource allocation based on deep reinforcement learning is designed.Simulation results show that our proposed scheme outperforms existing strategies in obtaining status information updates in a timely manner.