Research And Realization Of Congestion Control For Streaming Transportation Of MPEG-4 Video

With the rapid development of Internet and multimedia technology, the streaming transportation of MPEG-4 video has become a research hot spot. Congestion control is essential for transporting streaming video, it is crucial for the robustness and stability of the Internet. Video applications typically have delay and loss requirements, which cannot be adequately supported by the current Internet. A dominant portion of today's Internet traffic is TCP-based. Therefore, designing a TCP-friendly congestion control mechamnism suitable for transporting MPEG-4 video becomes important.The thesis first introduces the research background, development nowadays, principle of RTP/RTCP and the features of MPEG-4 video coding standard. Then it puts forward an end-to-end transport architecture based on RTP for MPEG-4 video streaming over the Internet and studies the playload type and packetization algorithm of MPEG-4 video bit-streams to achieve both efficiency and robustness for Internet MPEG-4 video.Afterwards, the thesis makes researches on the congestion control based on AIMD and TCP throughput model, improves the schemes, and presents our own congestion control mechanism.The thesis also introduces the platform of our experiments, the method and process of the implement of our congestion control schemes.Finally, the thesis summarizes our work and points out the research direction in future.The main contributions of the thesis are:(1)We present an end-to-end transportarchitecture using the RTP/UDP/IP protocol stack and employ an efficient and robust packetization algorithm for MPEG-4 video bit-streams at the sync layer for Internet transport. (2)We study the congestion control mechanism based on AIMD algorithm, and make improvement in order to reduce the oscillation of transimition rate due to tremendous contrast of packet loss ratio caused by dynamical change of the network load. We introduce two thresholds of the packet loss ratio to divide the network load condition into three status: congest, normal, idle. We import a lowpass filter to smooth the loss ratio, adjust sending rate according to the smoothed result. We present an adaptive low threshold scheme regulating the low threshold according to the history record of sending rate adjustment to utilize efficiently the network bandwidth, and reduce the oscillation of sending rate. We find that such a rate adjustment can shorten congestion period and reduce packet loss.(3)We propose a model-based congestion control scheme named MBCC. We chose suitable TCP throughput model to estimate the available bandwidth correctly, using the estimated round trip time and packet loss ratio for the next time interval as parameters of the model to achive the accuracy of estimated network bandwidth. As the observed losses and round trip time vary very dynamically, adjust the sending rate equivalent to the amount of TCP throughput may result in a rather fluctuant sending rate. So we present a rate adjustment like TCP congestion control based on AIMD, which increases its sending rate by an additiveinereease rate.(4)We implement our proposed algorithms on network simulator and perform a simulation study on transporting MPEG-4 video over the Internet. Simulation results showour AIMD congestion control achieves efficient utilization of network resources and TCP-friendliness. Our model-based congestion control mechanism can optimally track the network conditions, and has two good features: "TCP-friendliness" and "rate smoothness".