On Active Queue Management Algorithms Based On Router

From the Local Area Network(LAN)to the worldwide Internet,the scale of the network has increased larger and larger,which causes further the explosive growth in the data-flow.Since the network has been an integral part of our daily life,it plays more and more important role and provides a lot of convenience to society.On the other hand,people are bringing much even more requirements for network in the future.Among the unsatisfied requirements,congestion is a typical problem to be solved.Congestion is caused when the receiver cannot handle the data from the transmitter,and thus the data pile up in the bottleneck.According to many applications,congestions tend to occur in the routers of low capacities.In fact,the essential reason for congestion is the assigned unbalance in the resources and the data source of low link bandwidth,because the limi-tation of the physical structure of network results in slow data sending and leads to more jams.Thus how to guarantee the stability of the network and to avoid congestion have become a hot topic in research.At the same time,the active queue management based on router is a useful method which is widely used.Compared to the traditional Drop Tail method(Passive Queue Management,i.e.,drop packets when the queue is full),the Ac-tive Queue Management(AQM)method(i.e.,drop packets actively to adjust the queue length before the queue is full)is more effective.The AQM method can reduce the round trip delay,increase the utilization ration of the communication link,improve the Quality of Service(Qos)effectively.Therefore it has important theoretical and applicable value.Motivated by the above observations,this dissertation aims to deal with the drawbacks in existing results such as the uncontrolled sending rate directly in source node,the sensitive-ness to parameters,and the dependence on mathematical model,a series of approaches are presented to settle the aforementioned problems.The main content is organized as follows:First,the reason of congestion and the background and meaning of study are dis-cussed.The development and existing main research methods of active queue manage-ment based on router are reviewed in details.Then an improved method called active window management is introduced to over-come the disadvantages of source algorithm and active queue management.Combined with the advantages of the two methods together,an improved system model with less non-linear characteristic is presented,which is useful in controller design,and the condi-tions to guarantee the stability of the system to be satisfied are given.Moreover,if some parameters of the system model are unknown,an adaptive observer is proposed to for the controller design based on the router as well.Further,the Minimax methods for improving the robustness are presented to over-come the unknown disturbances in linearized system and nonlinear system respectively.As for the unknown disturbance in the linearized system,the Minimax algorithm is used to constrain the worst effect caused by the disturbance to the system.A robust controller is designed to make the system stable even when the worst effect happened.As for the nonlinear system,a controller is designed according to the characters of the nonlinear sys-tem,with combining Backstepping and Minimax method.It can make the system stable as well.Simulation examples verify the effectiveness of the proposed method.And then an improved switched gain based data-driven Iterative Learning Controller(ILC)is introduced for nonlinear model.Without knowing mathematical-model exactly,an switched gain based ILC algorithm is given,in which a switch gains PD-PD controller is designed.And the convergence rate of the system is guaranteed.It also prevents the instability of system due to increasing the gain constantly.The simulation examples verify the effectiveness of the method proposed.Then for multiple routers network,a cooperative controller of routers is introduced,in which each router is considered as a node.Some nodes in the network are coupled.To the linear and nonlinear growth couples,different observers,cooperative control algo-rithm and switch mechanism are used to control coupled node to achieve corresponding performance,and keep the communication stable.The merit and effectiveness of this algorithm are verified by simulation results.Finally,the results of the dissertation are summarized and further research topics are pointed out.