Robust Queue Management over a Link with varying Capacity

Active Queue Management (AQM) is the process of signaling Transmission Control Protocol (TCP) sources from core routers with the objective of managing queue utilization and delay. AQM methods can be classified into two main classes, namely, non-control theoretic-Random Early Detection(RED) and its variants and control theoretic-Proportional (P), Proportional Integral (PI) controllers. Shortcomings from RED motivated alternative AQM schemes which amount to classical P and PI control. But the robustness of these algorithms is rarely addressed in wired congestion control studies. In wireless networks, where the available link bandwidth is time varying, it is important to study the robustness of the congestion control algorithms against the capacity disturbance. Control theoretic approach is developed for the design of AQM Controller using Linear Quadratic Gaussian (LQG) theory and H∞ control technique. The aim of the control law is to minimize the perturbations in queue (from the set value) at the congested node buffer by choosing minimum cost input i.e. probability of drop. Result showing impact of probability of drop on packet drop and capacity utilization is given.