Parameter uncertainties and variations as well as input or state delays are crucial factors in the analysis and control of packet
flows using the Internet Protocol (IP) suite, of which the Transmission Control Protocol (TCP) is one of the core protocols.
The main TCP task is to provide reliable, order and error–checked delivery of streams of octets between programs running on
interconnected computers. In this paper an Active Queue Management (AQM) robust control strategy for TCP data transfer is
proposed. It is based on the second–order model with delayed input obtained from the linearization of an efficient and commonly
used nonlinear fluid–based model. To guarantee a satisfactory behavior of the adopted feedback control structure, a fractional–order
PI controller is proposed. In particular, a method for determining the parameter regions where a P I λ controller ensures a given
modulus margin (inverse of the H
∞ norm of the sensitivity function) is presented. The suggested graphically–based technique is
simple to apply and ensures robustness against disturbances and unmodelled dynamics.
