State-Space Models for Control and Identification
Abstract
In the last two decades, design and performance evaluation of efficient congestion control methods for packet-switching computer communication networks has emerged as a major engineering challenge. When congestion control is considered in the framework of control theory, communication networks are viewed as dynamical systems whose inputs and outputs are rates of data transmission.
Most attempts to formulate congestion control problems in the language of control engineering have focused on an input-output approach, where network elements such as buffers and transmission lines are modeled by differential, delay-differential or difference equations. If these equations are linear or can be realistically linearized around some equilibrium point or trajectory, standard transfer function techniques can be applied to perform controller design and to evaluate open and closed-loop performance criteria such as stability, robustness margins, transient response or disturbance attenuation.