Research On Intelligent Congestion Control In Complex Network

For complex,diverse and heterogeneous networks,the research of stable,efficient and intelligent congestion control is an important and challenging issue.Complex network model based on multi-layer generalized operator is proposed under the guidance of large systems control theory and method of intelligent model.Then,on the basis of multiple generalized operator model,the complex network of intelligent congestion control is researched from the perspective of horizontal and vertical variable granularity system.The various aspects about the development of network technology,the deepening of application and the popularization of users are continuously promoted and integrated,which brings the era of multi-object,heterogeneous and hybrid network.The complex network aggregates the characteristics of network in technology,structure and data.The intelligent congestion control is researched from two representative complex network systems,end-to-end and non-end-to-end.As a complex and large system,the study of congestion control is not an independent and autonomous process,whether in the network source based,link or routing strategy of mobile nodes.Instead,all objects and processes should collaborate to achieve the goal of collaborative intelligence and control optimization.The main contents and highlights of this dissertation are as follows:(1)The system architecture is studyed which is combined with the multiple generalized operator model and the intelligent model,Then,the concept of complex network model of multi-layer generalized operator is defined.According to the connotation of generalized modeling from large system control theory,the paper discusses the constraint conditions which are constructed the generalized operator model of complex networks.Referring to the idea and method of cloud model,the model of assessment method is proposed and evaluation procedures are designed.It is applied to the Internet modeling evaluation,and the evaluation process is given to verify the effectiveness and rationality of the method.(2)The problems existing in source based congestion control algorithm are analyzed and discussed,such as high bit error rate,low bandwidth,asymmetric link,long time delay in hybrid heterogeneous complex network.On the basis of analysis of the uncertainty in the parameter settings,the ambiguity of congestion degree and how to handle the diversity of packet loss,an adaptive source based congestion control algorithm based on bandwidth estimation is proposed.The link bandwidth and its volatility can be accurately estimated in real time for different network conditions,and we initially realized a certain degree of decoupling of some key indicators of packet loss,delay and congestion control parameters in network transmission.The bandwidth utilization factor is used to reflect the network congestion state and perform network congestion control.The experimental results show that the proposed algorithm can perform real-time congestion control for different network conditions,and the throughput,packet loss tolerance and delay are significantly improved compared with the existing congestion control algorithms.(3)The importance of link based congestion control is discussed,and in view of the performance instability caused by the linear relationship between queue length and packet loss setting in the traditional link control algorithm,as well as the optimization problem of variable parameters,a nonlinear link congestion control algorithm based on membership cloud theory is proposed by using the semi ascending cloud distribution model.The algorithm focuses on solving the nonlinear processing of packet loss rate function and parameter adaptive dynamic setting problem.The simulation results show that the performance of the algorithm in packet loss rate,average queue length,delay jitter,throughput all are better,and the global synchronization phenomenon of RED algorithm is also improved.(4)Applications and focus are constantly expanding about complex network without complete link between ends.Taking opportunity network as an typical subject,the congestion control algorithm is studied.Drawing on and expanding the research results of social network,the social attribute and association of nodes are mined under the deep learning model.Based on social awareness,an opportunity network routing strategy is proposed,in which node relationship and community cooperation are mined and applied reasonably.The algorithm not only combines the social attributes of nodes but also fully considers the dynamic evolution characteristics of sociality.According to the strength of social relations among nodes,the nodes are dynamically and adaptively divided into multiple communities,and the social attributes of nodes and the cooperation between communities jointly complete opportunistic routing and forwarding.The simulation results show that the algorithm can effectively improve the delivery rate,reduce the forwarding delay and the consumption of network resources.