The Congestion Control Algorithms Of Large-delay Networks

With the evolvement of Internet based on TCP/IP, the Scale, users and traffics of it have experienced an explosive growth since 90's. The network congestion has become more serious and complex due to the ever-increasing network application types. Congestion often results in decline of quality of service, while the network resource utilization is also affected seriously. The congestion control 1 is always a hot spot in the field of network research.The active queue management (AQM) acting on the intermediate nodes is able to control the queue length effectively on the basis of the high throughput of the routers. However, most AQM algorithms in the design and verification process do not fully consider the impact of the large-delay network. So that the AQM algorithms of large-delay netwoks has become a research direction.Considering large-delay of networks and uncernty of networks model, based on the analysis of congestion control mechanism of TCP/IP in control theory, the research work is summarized as follows:Considering the controls of PI and PID controller are not effective, and the weakness of depending on precise model of the Smith controller. An improved Smith controller as AQM algorithm is proposed based on the Smith controller. The simulation results of the algorithm under different instances show the superiority of the improved Smith controller.Considering the weakness of depending on precise model of the Smith controller, and the controls of improved Smith controller is imperfact, an active queue management algorithm is proposed based on the combining of self-learning single neuron PID controller with Smith controller. The defect of depending on precision model of Smith controller can be overcame by self-learning single neuron controller, which may work well without the precise model of real process. The single neuron PID Smith controller is robust to dynamic network parameters and suitable for networks with large-delay.The conclusion is drawn for the whole thesis, and the further research aspect is put forward.