Active Queue Management of TCP Flows with Self-scheduled Linear Parameter Varying Controllers

Abstract

Control-theoretic approaches to Active Queue Management (AQM) are typically based on linearizations of fluid flow models around design conditions. These conditions depend on the Round Trip Time (RTT), and the AQM performance is known to degrade if RTT values during actual operation depart substantially from design values. To overcome this difficulty a self-scheduled LPV controller for AQM is considered in this paper, where the controller is modified in real-time based on RTT. Simulations show that the self-scheduled LPV controller has good performance for both constant and time-varying RTTs, and outperforms two other common control-theoretic approaches to AQM.