Research On Scalable Video Streaming Mechanism Of Wireless NDN Based On Reinforcement Learning

With the rapid development of network technology and the mass popularization of mobile devices,content-intensive applications have developed rapidly,and video traffic has accounted for a large proportion of the total Internet traffic.Users’ pursuit of high-quality viewing experience is not only of practical and economic significance to video providers,but also due to its content-intensive nature,video content has the characteris-tics of large data volume and high bandwidth requirements,which brings huge challenges to the network carrying video stream data transmission..In addition,due to its low access cost,wireless networks are currently the main access method for a large number of users to request video content.Therefore,the research work of this paper is how to optimize the transmission efficiency of video streaming in wireless network.Due to its host-to-host architecture design,the TCP/IP network suffers from low hardware storage and bandwidth utilization during video streaming.As a content-centric network architecture,NDN has good support for multicast and in-network caching functions,which can significantly reduce the excessive duplicate content carried by the backbone network and further optimize users’ QoE.In addition,this paper uses SVC to encode and decode the video to improve the flexibility of video download.Therefore,this paper designs a scalable video streaming adaptive mechanism for wire-less NDN networks.The main research work is as follows:1)Due to the high dynamics of the NDN network,it is difficult for the traditional bitrate adaptation algorithm to estimate the bandwidth ef-fectively,resulting in the inability to adapt to network fluctuations in time to adjust the download strategy.Moreover,there is the problem of the huge download decision space introduced by SVC.Thus,this paper proposes an adaptive algorithm for wireless NDN video streaming based on reinforcement learning.NVRSA makes a targeted model design for the above-mentioned problems,it adjusts the online agent strategy that interacts with the environment through the offline training algorithm under the condition that it is not based on the assumption of the network environment,and obtains the best decision strategy for video chunk download.2)An NDN multicast group is formed based on multiple consumers requesting the same content.When multiple STAs in a wireless NDN network request popular video content,a multicast group will be frequently formed when channel competition occurs,further aggravating the dynamics of the wireless NDN network.Therefore,on the basis of NVRSA algorithm,this paper proposes a wireless NDN multicast rate selection algorithm NMRA based on reinforcement learning.Aiming at the lack of feedback in traditional wireless multicast schemes,NVRSA extends the interest and data packets of NDN,and uses a multi-agent-based offline training algorithm to maximize the overall transmission efficiency of data in wireless networks.3)Aiming at simulation problem for NDN video streaming.This paper reconstructs the code of the AMuSt video stream simulation project,and implements and provides a video stream simulation module DAS-NDN based on the NS-3 simulation platform.The experimental results show that compared with other bit rate adap-tive algorithms,the NVRSA algorithm can improve the video quality by about 28% and the user QoE by about 48% in both static and dynamic net-work environments for a single STA scenario.On this basis,for the multi-STA scenario,compared with other multicast rate selection algorithms,the NMRA algorithm can improve the video block download speed by about 62%,the video quality by about 15%,and the user QoE by 54%.Com-pared with the original video transmission scheme,the video transmission scheme combining the two algorithms proposed in this paper can improve the user QoE by 79%.