| We summarize in this thesis our progress on applying advanced feedback control theory to address design problems in communication networks. Two major directions are explored. Our attention is first focused on designing Active Queue Management (AQM) schemes for controlling Internet congestion in TCP/IP networks. We then proceed to decentralized constrained control of a class of large-scale networks (not restricted to TCP/IP networks).; In Chapter III, constructive feedback control designs for AQM routers are proposed, where our emphasis is on the output-feedback control scheme. The significance of using output-feedback is that the control strategy does not rely on measuring end-host window size, to reduce transmission overheads and implementation complexity. Only the easily accessible information of bottleneck queue length is measured for feedback control. With the help of observer-based backstepping design technique, we obtain the first output-feedback solution to Internet congestion control for the nonlinear network model. In contrast, most previous congestion control schemes are based on linear models (linearization) and on applying frequency domain approaches or linear PID control ideas (see [7][39][49][75] [93][106][110][12][48][44][115]).; To further demonstrate that tools from nonlinear system theory can play an important role in tackling "hard nonlinearities" and "unknown disturbances" in large-scale networks, we study in Chapter IV the queue regulation of a class of networks modelled by interconnected nodes (network resources such as routers, switches or wireless mobile units), with control input and state constraints. This issue is pertinent to the quality of service for premium Internet traffic [90][89].; The discussion for the large-scale networks is extended to the more practical scenario with propagation delays and to the case with unknown traffic bounds, both referred to in Chapter V. By choosing an appropriate Lyapunov-Krasovskii functional [82][43] and by applying Lyapunov's Second Method, we show that the achieved asymptotic stability is robust against small delays. The novel adaptive control scheme tracks uncertain upper bounds of traffic and achieves asymptotic queue regulation under control capacity constraints. Our conclusions and future directions for research are discussed in Chapter VI. |
