Cross-Layer Transmission Optimization Of Adaptive Streaming Media Over Wireless Networks

With the evolution of mobile networks and the popularization of portable smart devices,users’ demands for video streaming media services are growing exponentially,and video streaming media services have become the largest part of the traffic in wireless networks.Users’ demands for high-bandwidth services such as high-definition video and new streaming media are increasing day by day.Obviously,existing network resources cannot support multiple users to perform smooth high-bandwidth streaming media services.In addition,problems such as changes in the receiving environment and inter-cell handovers caused by user mobility will lead to violent fluctuations in the bandwidth available to users,thereby affecting service experience.A common practice is to use adaptive streaming media transmission technology,however,existing adaptive streaming media transmission algorithms rely on the accuracy of network bandwidth estimation and cannot adapt to changing wireless network channels.Therefore,through the idea of cross-layer optimization,the physical layer information of the wireless network can be introduced to assist the adaptive streaming media transmission technology to better adapt to the wireless network channel.In addition,different users usually have different viewing experience preferences for video streaming media services.How to dynamically adjust the transmission strategy of streaming media services according to different users’ experience preferences is also the focus of this thesis.In response to the above problems,based on DASH(Dynamic Adaptive Streaming over HTTP),this thesis introduces the physical layer information of the wireless network in a cross-layer optimization way to assist the network bandwidth estimation,and combines deep reinforcement learning technology to design two wireless network adaptive streams.The cross-layer transmission scheme of media can make full use of the bandwidth resources provided by the wireless network and provide users with a higher viewing experience.The main work of this thesis is divided into the following two parts:The first part,for vehicle or subway users,their high mobility will lead to frequent inter-cell handovers and sudden changes in the receiving environment,which will eventually lead to violent fluctuations in network bandwidth,making network bandwidth unpredictable,and ultimately leading to adaptive streaming media applications.In order to solve this problem,by introducing the physical layer and user location information in the wireless network,this thesis predicts the available network resources of the user in the future time period in the fom of cell clusters,and proposes a cross-layer adaptation based on cell clusters.The streaming media transmission scheme assists users to perceive the network channel conditions more accurately,realizes accurate tracking of network bandwidth,and better implements adaptive streaming media services in wireless networks.The second part,the existing adaptive bit rate selection algorithm depends on the accuracy of network bandwidth estimation and the user’s viewing experience preference.To solve this problem,this thesis firstly introduces the physical layer information of the wireless network to improve the bandwidth estimation in the wireless network.and using reinforcement learning technology to dynamically adjust the adaptive bit rate selection algorithm according to the preferences of different users,a cross-layer transmission scheme of adaptive streaming media in wireless networks based on reinforcement learning is proposed,so as to improve the performance of wireless networks.Compared with the best adaptive streaming algorithm at this stage,the video viewing experience of users with different preferences improves the viewing experience by 5%to 8%.