On Active Queue Management Algorithms In TCP/IP Network Congestion Control

Nowdays, with the rapid development of science and technology, the appearance of varied network applications bring the growth of users. The network congestion has become an important problem, which restricts the development and application of networks. Active queue management (AQM) is a kind of congestion control mechanism based on router. Nowadays AQM is extensively studied due to good effect for constraining congestion. But TCP network is a complex large system with respect to the nature of nonlinearity and uncertainty, which requires a kind of more robust AQM algorithm in order to obtain better congestion control effect.Some active queue management algorithms are presented based on robust control theories, intelligence control and optimal control. We make further to study on the design of active queue management and stability analysis. The main research works and conclusions are as follows:First, two novel active queue management (AQM) algorithms are presented based on sliding mode control theory for the problem of congestion control in TCP (transmission control protocol) networks. Considering TCP linear dynamic systems with state delay and input delay, through a particular linear transformation, the original uncertain time-delay system was first transformed into a delay-free system. Then based on the transformed system, the optimal sliding hyperplane is designed; a new reaching condition is proposed using the method of norm-bounded. And the stability of the system is rigorously theoretic proved; for the linear dynamic systems with input time-delay and unmatched uncertainties, an asymptotically stable sliding surface is designed by linear matrix inequality (LMI). The sufficient condition is obtained for the existence of the sliding surface. The effect of uncertainties and state delay is compensated by the designs of sliding surface and controller. The proposed sliding mode AQM controllers that satisfy the reaching condition can effectively constrain oscillation of the queue length in router and force the queue length to converge to the desired value quickly.Then, considering the effect of uncertainties, time delays and nonlinear to networks together, an active queue management (AQM) algorithm is investigated based on adaptive fuzzy sliding mode control (AFSMC). Sliding surface is constructed based on a Lyapunov-Krasovskill method for the particular network model, a sufficient condition is proposed for robust asymptotic stability of the system in terms of linear matrix inequality(LMI); the corresponding reaching law is designed, which can drive the state trajectory of system onto the sliding surface within limited time. The simulation results show that it can track queue length very quickly under various network conditions.Two observer-based active queue management (AQM) schemes are proposed because the system state is not fully measurable. In the present of input saturation, the proposed scheme for the uncertain linear system with time-delay is discussed. Using the linear matrix inequality approach, a sufficient existence condition of the observer-based controller is obtained; a robust active queue management (AQM) scheme based on H-infinty theory is presented for considering the inaccuracy, uncertainty and external diaturbace. The simulation results show that this method enables the buffer queue quickly converges to the expected queue length and has the strong robustness for the loading perturbation and the parameter change.Considering the nonlinear term and time-delay in the network systems, the state estimation problem gets more difficult. The traditional observer can not solve the problem. For this problem, the sliding mode observer is designed. The feedback compensatin term is introduced to counteract the affection of the time-delay and uncertainty terms, so the observers’ problem is realized for the uncertain time-delay systems. Then, an observer-based sliding mode controller is presented using the estimation state. Simulation results demonstrate that the proposed observer-based sliding mode controller can obviously improve the performance of congestion control for queue length in router.Takagi-Sugeno (T-S) fuzzy model can suitably representate a nonlinear system, an observer-based T-S fuzzy controller is designed. A T-S fuzzy modeling is done for the nonlinear TCP/IP network congestion control system. The performance of network congestion control system is improved by choosing appropriate fuzzy rules and membership functions, and the stability of the system is rigorously theoretic proved. Simulation results in different scenarios demonstrate that the proposed controllers have good stability and robustness with respect to the uncertainties of the number of active TCP sessions, link capacity and the round-trip time (RTT).An acticve queue management algorithm based on global sliding mode control is proposed for the nonlinear network systems. The scheme can remove the reaching phase from sliding mode control, and to guarantee the system robustness during the whole control process. The algorithm shows its good transient and steady state responses in TCP/IP networks, such as modeling uncertainties, fluctuation of time-varying parameters and network jittering due to non-TCP sessions. Simulation results demonstrated that this method enables the queu length to converge to set value quickly and keeps the queue oscillation small.Lastly, the summary of the whole dissertation is given and the research directions in future are put forward.