On AQM Algorithm Based On Intelligent Sliding Mode Control

With the rapid development of the network and the proliferation of network users,the problem of network congestion has become more and more concerned.In recent years,short video,live video,double eleven online shopping and other online entertainment activities have entered the public life,which requires a good network to serve the public.Network congestion has become a very important issue.Traditional active queue management algorithms are difficult to provide high quality of service in today’s network environment.The combination of the active queue management mechanism and the congestion control of the TCP protocol is currently a major method for solving the congestion control problem of the TCP network.Sliding mode variable structure control is robust to nonlinear systems that deal with inaccurate model information.The combination of neural network and sliding mode variable structure control does not need to know all the accurate model information of the system,and has strong adaptability.In this paper,RBF neural network and sliding mode variable structure control are combined.and applied to the research of AQM algorithm.Two intelligent sliding mode variable structure AQM algorithms are proposed.MATLAB simulation shows that the algorithm has better gradual in various network environments.Nearly stability and robustness,the main research contents of this paper are as follows:Firstly,the various symbolic functions in the sliding mode variable structure approach law are compared,and the alternative symbol function with the best effect is selected.The effect of using neural networks to approximate nonlinear functions is studied.Secondly,for the nonlinear TCP dynamic model based on AIMD strategy,the influence of system parameter perturbation and UDP stream interference on the TCP model,and the part is not known accurately.The RBF neural network is used to approximate the characteristics of arbitrary nonlinear functions,and the function of the total uncertainties in the system is approximated.Then the sliding mode variable structure control design controller is adopted,and the single parameter learning method is adopted to speed up the solution.Finally,through the Lyapuno The stability of the function of the function is proved by the control law.The algorithm can make the actual queue length stabilize to the desired queue and has good robustness.Thirdly,for the nonlinear TCP dynamic model based on AIMD strategy,physical line faults,equipment faults,routing errors,etc.all have an impact on the packet loss rate.The perturbation of the packet loss rate is introduced into the model to make the model more realistic.For the packet loss rate disturbance,the neural network approximation is adopted,then the sliding mode variable structure control is adopted,and the controller is designed to obtain the control law through adaptive solution.The algorithm can make the actual queue stable without oscillation after considering the model that is more in line with the actual situation.The queue is expected to reduce the actual queue convergence time and has good robustness.Finally,the research work and main results of this paper are summarized,and the research direction of the next active cluster management algorithm based on intelligent sliding mode is prospected.