Research On Congestion Control For Computer Networks Based On Sliding Mode Control Theory

The Internet protocol architecture is based on a connectionless end-to-end packet service using the IP protocols, and provides only best effort service. However with the technology development of high speed networks and multimedia, integrated service requirement including multimedia communication is required more and more. DiffServ Network is a QoS service system with strong expansibility. It has syncretized best effort service and proposed premium service, ordinary service and other different level services. No matter what the traditional TCP/IP Network and DiffServ Network, designing a simple and efficient congestion control algorithm becomes an urgently problem in network management. As an important module in high-speed router, active queue management (AQM) has been got more and more recognitions in recent years. Implement of AQM scheme in high speed router is in order to provide small packet drops, high link utilization and low queue delay.In sliding mode control system, the sliding mode possesses of strong robustness, which is very suitable for complex time-varying network system. So in this dissertation, some AQM algorithms are proposed based on sliding mode control theory in order to solve the problem of congestion control for networks. The main innovative contributions of this dissertation are summarized as follows.The dissertation starts by TCP Vegas model, and the control theoretical based AQM controller has been studied. The model is linearized to derive a state space representation. The AQM controller is designed based on sliding mode control, which is reachable in finite time. The controller can make the queue length converge to desired value quickly. And both the transient state and stable state are ideal.A model combined with proportional Kelly scheme and control theory is proposed. Considering the integrated performance of both source and link, the linear and terminal sliding active queue management (AQM) controllers are designed. Especially in the terminal sliding AQM algorithm, a special nonlinear terminal sliding surface is designed in order to force queue length in router to reach the desired value in finite time. The upper bound of the time is also obtained. Meanwhile, an improved exponential reaching law is used to design the controller.Based on nonlinear fluid flow considerations and the Integrated Dynamic Congestion Control (IDCC) scheme for DiffServ networks, a fuzzy sliding mode controller and a sliding mode controller with input delay compensation are proposed for premium traffic and ordinary traffic separately. With the fuzzy control introduced, the arrival time is shortened and the jitter from sliding mode control is weakened. This method makes the control signal smoother, and the system performance be much better improved.Considering that Takagi-Sugeno (T-S) fuzzy model can suitably represent a nonlinear system, the sliding mode control and T-S fuzzy model are combined to carry out congestion control together. A T-S fuzzy modeling is done for the nonlinear DiffServ Network congestion control system. The performance of network congestion control system is improved by choosing appropriate fuzzy rules and membership functions. The queue length in router can converge to the desired value quickly and keep low oscillation in steady state. So link utilization is increased and the rate of packet loss is decreased. An asymptotically stable sliding surface is designed by LMI in order to effectively compensate for the effect of uncertainties and time-varying delays.Two important services of DiffServ Network are to be analyzed and controlled, one is premium traffic service and the other is ordinary traffic service. A nonlinear fluid flow model is used to design a new adaptive sliding mode controller. The model is transformed into a parametric strict feedback form, and then use backstepping procedure is used to complete the design. And by adding sliding mode, unknown system dynamics and uncertain disturbances are effectively constrained. Ideal buffer queue length effect is achieved. In order to realize the integrated dynamic congestion control strategy, a leader-follower control scheme is proposed. Based on this strategy, three kinds of second order sliding mode controllers are designed to regulate the buffer queue length for DiffServ Network. With respect to standard sliding mode control, second order sliding mode technique shows the same properties of robustness to model uncertainties and considerable simplification of the model used in the design. Apart from the robustness feature, the proposed second order sliding mode control laws have the advantage of being continuous, thus eliminate the chattering effect and being more acceptable in application.Finally, the conclusions of the dissertation are drawn and further research direction is put forward.