| In recent years, with the rapid development of wireless communication technology, various wireless communication system emerge in endlessly to provide users with heterogeneous network environments, including the WPAN (such as Bluetooth), the WLAN (such as Wi-Fi), the WMAN (such as WiMAX), the public mobile network (such as 2G,3G), satellite networks, Ad Hoc networks and wireless sensor networks, etc. Although these wireless networks provide users with a variety of communication mode, access method and ubiquitous access service, it is still necessary to fully realize the organic integration of heterogeneous wireless network technology to realize the self-organization, adaptiveness and end-to-end QoS guarantee. Communication service over heterogeneous networks is an inexorable trend. The "best-effort" service of traditional IP network cannot satisfy the needs of real-time multimedia traffic. Wireless channel is time-variant, susceptible and inefficient, and these disadvantages bring great challenge for wireless Internet providing QoS guarantees to multimedia transmission. The techniques of QoS exist in each layer of networks, including reasonable control mechanism and algorithm in wireless link and MAC/PHY layer, suitable transport protocol for multimedia transmission, and relevant agreement in application layer etc. In this paper, the problem of IEEE 802.11 wireless network in supporting video transmission with QoS is systematically analyzed. The research object is the wireless local area network (WLAN) and related wireless networks, and the research target is to enhance the QoS guarantee for video transmission. The emphasis of the study is on the performance optimization of MAC layer protocols and end-to-end optimization algorithm of transport layer protocols. The research of the thesis mainly includes the following aspects:1) The research on IEEE 802.11e for video transmission with QoS guaranteeIEEE 802.11e is based on IEEE 802.11 and it aims at providing QoS guarantee for realtime multimedia traffic. It introduces a novel concept named TXOP (Transmission Opportunity), which can allow a station transmitting multiple continous packets in one access opportunity, thus the chance of collision is avoided in next contention. This greatly improves the network efficiency and reduces the transmission latency. But the existing admission control rule in QAP (QoS-capable Access Point) dose not consider the VBR (Variable Bit Rate) video stream with great fluctuation of bitrate, it only allocates equal TXOPs to flows with equal average bitrate. When a VBR flow contends with a CBR (Constant Bit Rate) flow with the same priority, it is difficult for VBR flow to obtain the granted service quality for its feature of bursty arrival. Aiming at this problem, we propose an improved mechanism in TXOP allocation. A SVM (Support Vector Machine) model is used to learn and predict the frame size of VBR video. The model is first trained by previously arrived frames, and then is used to predict the forthcoming frames by the QAP. The QAP can adaptively allocate the needed TXOP according to the predicted frame size to be transmited. In this way, the burst packets won't be delayed for without enough TXOP. The experiment results show that SVM model can predict the VBR frame size accurately, and the adaptive TXOP allocation and transmission scheme can meet the delay and throughput demand of VBR stream, and meanwhile the packet loss rate of the system is decreased.2) An end-to-end loss differentiation mechanism in wired-cum-wireless networksIn a hybrid wired-cum-wireless network environment, congestion may not be the only reason of packet loss as in wired networks. Wireless link errors may be an important reason. Traditional congestion control protocol, as exemplified by TCP-friendly rate control (TFRC), adjusts its transmission rate in response to the total rate of loss events and doesn't discriminate the losses due to wireless link error or congestion, and eventually this behavior can compromise the throughout performance of networks. Therefore, differentiating the cause of packet loss is important to help transport protocols take congestion control actions only on congestion-induced losses. In this paper, an end-to-end loss differentiation mechanism is proposed to improve the transmission performance of TFRC protocol. Our key design is to introduce the SVM (Support Vector Machine), an outstanding machine learning algorithm in pattern recognition, into the network domain to perform multi-metric joint loss differentiation. It uses end-to-end indicators, such as the relative one-way trip time and the inter-arrival time of packets fore-and-aft the losses, as input features of the SVM, and doesn't require support from intermediate network apparatus. Simulations are carried out to evaluate the loss differentiation algorithm under various network configurations including different number of competing flows, wireless loss rate, and queue buffer size. Results manifest that our classifier is effective under most scenarios, and that the performance is superior to the ZBS (ZigZag, mBiaz and Spike) scheme.3) An end-to-end loss differentiation mechanism in MANETThe loss reason in MANET (Mobile Ad hoc NETwork) is analyzed and it is pointed out that packet loss caused by congestion accounts for only a small part. In addition, un-reliable wireless channel and frequent route change caused by mobility create considerable losses, which bring about serious influence to network performance. Aiming at this problem, we propose to use SVM to differentiate the three loss reasons:wireless random errors, route change losses, and congestion losses. The detailed procedure of feature extraction and selection are analyzed and discussed. On the basis of loss differentiation, the TFRC protocol is improved correspondingly. The sender can take different control strategy according to different loss reasones from the feedback. Extensive simulation results demonstrate that the differentiation scheme performs effectively under most scenarios and the network throughput is obviously improved.
|
PDF Full Text Request