Design of parameter tunable robust controller for active queue management based on H∞ control theory

Abstract

The issue of congestion control in communication networks has been attractive in recent decades and is becoming more and more important and challenging. Active queue management (AQM) has been proved to be an efficient way to alleviate or control network congestion. The existing AQM algorithms suffer from parameters tuning and robustness. They thereby are incapable to adapt the dynamics of complex networks characterized by large delay, traffic burst and uncertain users. A robust controller for AQM based on modern H ∞ optimal control theory with parameters tuning is presented in this paper. Taken both robustness and closed loop performance into consideration, most desirable parameters can be regulated by the analytical formulas derived in this paper. Then we take an example for proportional–integral–differential (PID) to materialize the proposed controller. Our robust PID controller can be tuned by only one parameter, unlike traditional PID controllers which are tuned by three or more. Theoretical analysis shows that the robust controller can converge to the reference value at a steady state. Simulation results confirm our expectation that our PID controller for AQM outperforms the existing famous congestion controllers, such as RED and PI, on keeping queue size at the target value at routers. The most important feature is that the controller is robust to the network dynamics including the changes of types of traffic and traffic load.