Active Queue Management Algorithm Based On Control Methods

With the rapid development of Internet, network has been widely applied, while more and more serious congestion problems have been exposed. Network congestion control is one of key factors to ensure the stability of Internet since it can effectively improve network's performance. As a complementary method to end-to-end congestion control, the Active Queue Management (AQM) algorithm take effective preventive measures before the congestion occurs by dropping packets purposeful, in order to ensure the network quality of service.Active queue management algorithms for Internet congestion control with control theory is emphasized in this paper.Firstly, the traditional PID controller is designed based on a linearized model of TCP congestion control system, using the classically method in frequency domain of control theory. And an adaptive PID algorithm is proposed based on the AQM algorithms of static PID. Then we compare the performances of PI, PID and APID in NS.Secondly, a robust H_∞controller based on LMI is proposed, whose coefficient can be determined by the linear matrix inequality, via giving a robust stabilization theory of the linear state-delayed system. Simulation results indicate the proposed controller has good adapt ability under diverse network conditions, and always maintain the system operating in the vicinity of the equilibrium point so that the queue length can keep in the vicinity of the expected value, and avoid the network congestion efficiently.Thirdly, a combination fuzzy controller is proposed based on fuzzy sets and fuzzy system theory, which adjusts its packet drop probability by the queue length and packets arrival rate in the router. So the queue length can stay in a expected value.Finally, considering the parameter-uncertain and time-delay problem of the system, a new active queue algorithm called S-PID is proposed, which is combined by single neuron controller and improved smith predictor. The new algorithm has advantages of the compensation control and the self-learning ability, the self adaptability. Simulation proves that the algorithm has good performance of robustness and the adapt-ability, and is superior to the PID control.