Active Queue Management Algorithm Based On Fuzzy Control Theory Is Studied

With the development of network technology and the growth of the number of users, the issue of network congestion is getting more and more difficult to deal with for the flow control mechanism of TCP source, Active Queue Management (AQM) therefore has become a hot spot in the field of network congestion control, supplementing end-to-end algorithms. Due to the non-linear and time-varying nature of the network, classical AQM schemes with fixed parameters are incapable of meeting their performance requirements, bringing out the necessity of designing AQM algorithms that can adapt to changes in network condition better. In this paper, we propose some new AQM algorithms using fuzzy control, major results are listed as follows:(1) A PI AQM algorithm based on Takagi-Sugeno fuzzy model is proposed. Considering the non-linear and time-varying characteristics of network congestion control system, a non-linear model is constructed, which fits network congestion control system to a satisfiable level employing different stable state and locally linearized models with T-S fuzzy model. A novel AQM algorithm is designed using the PI controller based on this model.(2) An AQM algorithm based on T-S fuzzy model is proposed. A non-linear model of the TCP/AQM congestion control system is developed using T-S fuzzy model, and a novel active queue management algorithm is designed using state feedback control based on this model. The conditions of the system’s asymptotic stability are given, also given is the design methodology of the AQM controller’s parameters by using Linear Matrix Inequality.(3) An AQM algorithm based on queue length and packet arrival rate is proposed. Two fuzzy controllers respectively based on queue length and packet arrival rate are proposed, the two controllers are combined to form one single controller, and through the parameter w to adjust the speed controller. In addition, the parameter adaptive mechanism for the queue controller is designed, improving the robustness of the algorithm.NS2simulation results show the aforementioned algorithms can maintain the queue length around the given target, and has good stability and robustness.