Research On Active Queue Management Algorithm Based On Intelligent Control

In the past twenty years, the computer network has experienced explosive growth. The congestion problems are also serious day by day. The congestion control is important to guarantee the Internet robust. It is also a necessary prerequisite of other quality of service for the normal work. Therefore it becomes a hot topic currently. In the congestion control domain, AQM is a hot spot research direction. It can make a reasonable balance in the high throughput and low delay. From the point of view of control system, this is the behavior which governs system's overall performance. The control theory is a quite mature system theory. Some many methods may profit the congestion control, and enhance its performance.Firstly, the congestion control mechanism which at present often uses is analyzed, and the representative AQM algorithm is introduced, then the feasibility of the introduction of intelligent control to congestion control is analyzed in detail.Secondly, on the basis of analysis and research on PI controller in-depth, AFPI algorithm is proposed by combining PI controller with fuzzy control, and has been given the detailed elaboration from the three aspects of fuzzy, the fuzzy inference, and the clear.Thirdly, the reinforcement learning technology is researched. Not only the technology that is combined the temporal difference technology with double gradient descent technology is made use of, but also the thought that congestion control signal first transmits is led into. Then, PRL algorithm is proposed. It is suitable to congestion control in the middle node.Finally, in the NS simulation platform, the experimental performance above algorithm from queue length stability, robustness, fairness and the RTTare simulated, and the experimental results are analyzed and discussed, then the feasibility of the AFPI algorithm and PRL algorithm are verified.