Intelligent Robust Aqm Algorithm And Its Fairness

With the rapid development of the internet information network, making network data stream sharp increase, the problems of network congestion become very serious, resulting in the decline in the indicators of business, low utilization of resources and unfair resource allocation and other issues. Congestion control algorithm is to solve the problem of the network congestion and ensure QoS, AQM algorithm based on the current router is an effective method to solve the network congestion problems. Based on the existing AQM algorithms, this paper proposes two intelligent robust control AQM algorithm focused on the robustness, stability and fairness of the network. The main work of this thesis are as follows:(1) About the large transmission delay problem in the system of TCP/AQM, an AQM algorithm based on neural network internal model control is proposed. Internal model control can be very good control for the system with large delay, which can remove the model error, and the neural network has a strong self-learning ability. The proposed AQM algorithm includes advantages of neural network and internal model control, and improve the system stability and robustness.(2) About the nonlinear feature of the TCP/AQM system, an AQM algorithm based on SVM-IMC is put forward. SVM algorithm is a new type of machine learning algorithm, which is based on the principle of structural risk mininization, and it can be a solution to limited number of high dimensional model structure of the sample, at the same time, the constructed model has good prediction performance. The proposed SVM-IMC AQM algorithm can improve the system stability and robustness.(3) The CHNN-IMC AQM algorithm which based on the combination of NN-IMC AQM algorithm and the flow identification mechanism of CHOKe is prposed to improve the stability and fairness. Meanwhile, the CHSVM-IMC AQM algorithm which based on the combination of SVM-IMC AQM algorithm and flow identification mechanism of CHOKe is proposed to improve the stability and fairness.