Research On Congestion Control For Wireless Networks

With the rapid development of the personal data communications, powerful portable data terminals and multimedia terminals have been widely used. So wireless networks will be an important component of Internet.Currently TCP protocol is dominated in the Internet, which is an end to end protocol that is designed to provide reliable network communication between nodes. It assumes that the quality of communication link is high, and link error rate is very low, only can congestions or a large delay cause packet loss. However, the wireless network is essentially different from the wired network, it has high error rate, bandwidth changes, and frequent movement of large time delay and other characteristics, thus the link quality is poor. So when traditional TCP protocol is directly applied to wireless networks, the performance of TCP will worsen sharply. In addition, the research on active queue management (AQM) algorithms deployed in the middle nodes of the wired network has achieved fruitful results, such as RED, Blue, and PID, etc. However, the research on AQM in the middle nodes of the wireless network has not fully started. Therefore, to maintain good wireless network throughput and high utilization, improving the existing congestion control protocol or proposing a new congestion control protocol is imminent.Today a typical representative of wireless networks end to end congestion control algorithms is TCP Westwood. It proposes a more accurate bandwidth estimation method, which estimates the available bandwidth of the current link by monitoring the ACK reception rate at the sending end. After the occurrence of packet loss, the sender resets slow start threshold (ssthresh) and congestion window (cwnd) according to the estimated available bandwidth, unlike TCP Reno halves the window blindly. So to some extent, it eliminates the impact of wireless loss. It significantly improves the throughput of the wireless network and has good fairness and interoperability with existing networks. But it can not distinguish between types of packet loss so that the congestion mechanism is called frequently, and still adoptes the TCP Reno's slow-start phase and congestion avoidance phase. The more well-known algorithms of Intermediate nodes algorithm are RED algorithm, Blue algorithm, PID algorithm and so on, but they are designed for the wired network and are not necessarily suitable for wireless networks.This paper describes the characteristics of wireless networks and the technical difficulties and deficiencies of congestion control algorithms deployed in wireless networks. Then discuss the direction of further research based on the analysis of a variety of enhanced algorithm. Then the above issues are studied, and put forward three kinds of improved congestion control algorithm for wireless networks, achieving to maintain a good wireless network throughput and high utilization. Results of this study are as follows:Firstly, TCP Westwood sets initial slow-start threshold of slow-start phase to too arbitrary values resulting in bandwidth utilization dropping, and so on adjustment of the window in congestion avoidance phase is too large resulting in the throughput of link is declining. The above issues have got the corresponding improvement. By the simulation results show that the bandwidth utilization and link throughput of slow-start phase and congestion avoidance phase has been significantly improved.Secondly, in high bit error rate wireless network environment TCP Westwood (for short TCPW) can not differentiate the wireless packet losses from the congestion packet losses. So, a new TCP scheme is proposed, called TCPW BJ, which is based on round-trip delay jitter loss differentiation and is a revision of TCP Westwood. It divides network status into different congestion levels by the measured round-trip delay jitter. According to congestion levels, it differentiates the wireless packet losses from the congestion packet losses and carried on the appropriate congestion control. The experiments show that in high bit error rate wireless networks TCPW BJ algorithm significantly improves bandwidth utilization and throughput, and maintains the fair and friendly behavior with respect to other TCP flows.Thirdly, after analyzing the characteristics of wireless links and the traditional single neuron PID controller, a new based on a single neuron adaptive PID controller is proposed and it is applied to active queue management algorithm for wireless networks. It overcomes two drawbacks of the traditional single neuron PID algorithm in the congestion control process:slow response and the impact of large delay. Simulation results show that compared with existing algorithms, the new algorithm can better adapt to the wireless network.Finally, this paper gives a summary and proposes some recommendations on how to improve the accuracy distinguish the packet loss and how to suitably combine end nodes and intermediate nodes, etc. This laid the foundation for future research.