Fuzzy Neural Network-based Congestion Control Algorithm For Wireless Networks

With the increasing number of wireless network terminals,the problem of network congestion has become increasingly prominent.As a mainstream link congestion control algorithm,active queue management(AQM)controls the amount of source-end transmission through link information,and effectively mitigates the link's congestion.However,the congestion control methods of traditional wired networks cannot satisfy the requirements of wireless networks because there are many differences between wireless networks and traditional wired networks.Therefore,we need to design a congestion control algorithm which is more suitable for wireless networks.In this thesis,the characteristics and disadvantages of traditional congestion control algorithms are considered.The traditional active queue management algorithm is improved which is based on the mathematical model of wireless local area network and combines with fuzzy logic control and neural network.Then the feasibility of the algorithm is verified by NS2 simulation software.The main work is as follows:1.In order to effectively study the wireless local area networks congestion control problem,this thesis proposes a TCP/AQM mathematical modeling suitable for wireless local area networks which is based on the traditional wired network TCP/AQM mathematical modeling and combines with the characteristics of the wireless local area network.After using local linearization to obtain the dynamic characteristics of the network,the principles of several active queue management algorithms are analyzed and applied to wireless local area networks.At last,simulation comparisons are performed on the topological model of wireless local area network through NS2 network simulation software,and the advantages and disadvantages of different algorithms is analyzed.2.Due to the problems of non-linearity,time-delay,and variable parameters in the wireless local area network,the wireless TCP/AQM mathematical modeling is not accurate enough.In addition,traditional PID congestion control also has difficult parameter tuning,the poor adaptive ability and poor adaptability to dynamic network environment.To solve these problems,this thesis designes an active queue management algorithm based on fuzzy control and PID neural network(FCAPIDNN).It can solve the problem that the learning rate of PID neural network can not be adjusted adaptively.Moreover,fuzzy control is used to optimize the accuracy of the learning rate adjustment and further improve the control accuracy.Finally,the NS2 network simulation software is used to simulate the wireless local area network topology model.The performance of the FCAPIDNN algorithm in terms of queue length,round-trip delay and packet loss rate is analyzed.The results show that the algorithm has good stability in wireless local area network.3.This thesis introduces a scheme that takes the packet arrival link rate as the input of the controller because the traditional network congestion control ignores the continuous state of network congestion.Then,an active queue management algorithm that improves single neuron gradient learning(ISNGL)is obtained.The algorithm uses gradient learning to dynamically adjust the network parameters and improves the convergence rate and stability.It proposes a new activation function with displacement parameters and an improved method with momentum adjustment for the weights.Finally,the NS2 network simulation software is used to simulate the wireless local area network topology model.The performance shows that the ISNGL algorithm has good congestion control capability in the wireless local area network.