【作者】 李超；

【导师】 徐少毅；

【作者基本信息】 北京交通大学 ， 电子与通信工程（专业学位）， 2019， 硕士

【摘要】 随着海量连接物联网时代的到来,在物联网场景下基于移动边缘计算的无线资源管理的研究已经引起了广泛的关注,这是未来无线通信发展的主要方向之一。在海量连接物联网的时代,必然会出现海量的数据,如果这些海量数据全部卸载到云核心网上,会导致云计算网络传输负载量急剧增加,造成较长的网络延迟,大大降低用户的体验。移动边缘计算(Mobile Edge Computing,MEC)作为未来5G网络提高用户体验度的关键技术,通过将计算能力下沉到移动边缘节点,有效地降低了网络传输延迟和能耗。但是移动边缘服务器的计算资源有限,同样不能将海量的数据都卸载到移动边缘服务器上。如果周围有空闲且计算资源丰富的终端设备,可以通过设备到设备(Device-to-Device,D2D)通信方式,和MEC服务器进行联合卸载和缓存来进一步提升蜂窝网络的计算和缓存能力。因此,本论文针对海量连接物联网场景下,如何有效提升移动边缘计算的无线资源管理的问题进行了研究,主要创新点如下:1)针对单MEC服务器多用户场景提出了一种任务卸载策略。针对计算密集型或延迟敏感型应用,提出空闲且具有计算资源的终端通过D2D通信和MEC服务器进行联合卸载,将系统模型建立成一个势博弈,通过合作-傀儡模型,得出势函数,随后利用最佳响应算法得出最优的卸载策略。仿真分析表明,所提出的基于最佳响应的MEC-D2D联合卸载算法有很好的收敛性,明显提升了传输的数据量并降低了任务能耗。2)针对单MEC服务器多用户场景提出了一种任务缓存策略。针对目前不同的流行视频流文件,提出空闲且具有计算资源的终端通过D2D通信帮助MEC服务器协作缓存,MEC服务器利用不同的流行度视频文件价格和设定的总报酬提出一个斯坦克尔伯格博弈(Stackelberg Game,SG)模型,建立领导者和追随者的双层博弈模型。3)同时,设立一定的激励机制,去激励D2D设备帮助MEC服务器进行协作缓存。在SG模型的基础上,利用逆向归纳分析法,对SG模型进行均衡分析,并分别对领导者MEC服务器的成本函数和追随者终端的成本函数进行求解,最后,通过仿真验证,本文提出的方案提高了 D2D设备参与协作缓存的积极性。更多还原

【Abstract】 With the development of the massive Internet of Things,research on radio resource management based on Mobile Edge Computing in the IoT scenario has attracted widespread attention,which is one of the main directions for the development of radio communication in the future.In the era of massive connectivity of the Internet of Things,massive amounts of data will inevitably occur.If these massive amounts of data are all offloaded to the cloud core network,the cloud computing network transmission load will increase dramatically,resulting in longer network delays and greatly reducing the user experience.Mobile Edge Computing(MEC),as a key technology to improve user experience in future 5G networks,effectively reduces network transmission delay and energy consumption by sinking computing power to mobile edge nodes.However,the mobile edge server has limited computing resources and cannot offload huge amounts of data to the mobile edge server.If there are idle and computationally rich terminal devices around,the device-to-device(D2D)communication mode and the MEC server can be jointly unloaded and cached to further improve the computing power of the cellular network.Therefore,this thesis studies how to effectively improve the radio resource management of mobile edge computing under the massive connected IoT scenario.The main innovations are as follows:1)A task offloading strategy is proposed for a single-MEC-server and multi-user scenario.For computationally intensive or delay-sensitive applications,it is proposed that this kind of tasks can be jointly offloaded to the idle terminal with abundant computing resources or the MEC server.This problem is formulated as a potential game,and the potential function is furtherly obtained through the cooperation-puppet model,,and then the best response is utilized to achieve the optimal solution.The algorithm yields an optimal offloading strategy.The simulation analysis shows that the proposed best response-based MEC-D2D joint unloading algorithm has good convergence,which significantly improves the amount of data transmitted and reduces the energy consumption of the task.2)A task caching strategy is proposed for a single MEC server multi-user scenario.For the current popular video stream files,it is proposed that the idle and computing resources terminal help the MEC server to co-cache through D2D communication.The MEC server uses a different popularity video file price and the set total reward to propose a Stackelberg Game.(Stackelberg Game,SG)model,establishing a two-layer game model of leaders and followers.3)At the same time,set up a certain incentive mechanism to encourage D2D devices to help MEC servers to collaborate and cache.Based on the SG,the inverse SG model is used to analyze the SG model,and the cost function of the leader MEC server and the cost function of the follower terminal are solved respectively.Finally,the scheme proposed in this paper is verified by simulation.It greatly stimulated the enthusiasm of D2D devices to participate in collaborative caching.更多还原