Modified Smith Predictor Controller for Network Congestion Control

Abstract

This paper presents a Modified Smith Predictor (MSP) controller to design a novel Active Queue Management (AQM) scheme. The controller decouples the setpoint response from the load response and shows a stable response for a Transmission Control Protocol(TCP) flow dynamics, which is highly oscillatory or unstable under practical conditions. The proposed control scheme enables to track the desired buffer level successfully and reject any load disturbances such as unresponsive traffic flows efficiently. The controller parameters are estimated by the use of standard forms of closed loop system response. Network simulator (ns-2) based simulation results show improved performances of the proposed controller compared to that of some popular schemes like Random Early Detection(RED), Proportional and Integral(PI), Linear Quadratic Regulator(LQR) etc.