Active queue management methods in computer communication networks based on pole placement and H-infinity optimal control.

摘要

This thesis studies the control of TCP traffic in the Internet nowadays. Both classical and modern control theory are used to design the controllers for the TCP/AQM system.; For the classical control theoretical approaches, we revisit the simple pole placement technique, and exploit this technique to propose four kinds of controllers for active queue management (AQM) in IP routers: the Proportional controller based on Pole Placement (the P_PP controller), the Proportional-plus-integral controller based on Pole Placement (the PI_PP controller), the Self Tuning Proportional controller based on Pole Placement (the ST_P_PP controller), and the Self-Tuning Proportional-plus-Integral controller based on Pole Placement (the ST_PI_PP controller). The transient performance indices---the damping ratio xi and the undamped natural frequency on, are appropriately chosen such that: (a) the transient response of the system is satisfied; (b) all the poles lie in the left-half of the s-plane to guarantee the stability of the system. The self-tuning controllers (the ST_P_PP and the ST_PI_PP) can maintain on-line tuning of the controller parameters so that good system performance can be ensured even under significant load changes.; This thesis also investigates the application of a robust modern control technique--- Hinfinity optimal control to the AQM controller. Two Hinfinity loop-shaping techniques, the S/U Mixed Sensitivity Problem (MSP) and the S/T/U MSP are applied to the controller design. Different weight functions and their impacts on the system performance are discussed and evaluated.; Finally, by modeling the uncertainties in the TCP/AQM system in terms of the uncertainty weight, this thesis also designs a robust AQM controller based on a modern control technique---robust mu analysis. The controller is synthesized by approximating the mu-optimal control problem with the Hinfinity S/T/U MSP. The NP (Nominal Performance), the RS (Robust Stability) and the RP (Robust Performance) of the system under the synthesized controller are analyzed using the mu analysis technique.