Research On Assessment And Prediction Technique Of End-to-end Situation In Cognitive Network

With the vigorous development of the information network and increasing types of networkapplications, users highly demand on the QoS requirement of business, but the “best effort” servicethat provided by existing IP network can not meet end-to-end QoS requirements of both user andbusiness. In response to this phenomenon, the academia introduces the concept of cognitivenetworks. A cognitive network is a network with a cognitive process that can perceive currentnetwork conditions, and then plan, decide, and act on those conditions. The network can learn fromthese adaptations and use them to make future decisions, all while taking into account end-to-endgoals.A cognitive network must provide better end-to-end performance than a non-cognitive network,whether a cognitive strategy is effective or not finally can be measured through monitoring theend-to-end QoS. In order to monitor the end-to-end QoS in cognitive network, this paperquantitative calculate end-to-end situation level. We primary research on the process of end-to-endsituational awareness in cognitive network, including the establishment of index system, end-to-endassessment, and the prediction of end-to-end trend. In this paper, we focus on three problems.Firstly, according to the similarity between level2layer of data fusion and end-to-endsituational awareness process, this paper puts forward a kind of cognitive model of networksituational awareness drawing on the Endsley data fusion model. In order to completely describe thecharacteristic of end-to-end situation, we establish a general and multi-level index system ofcognitive network situation information.Secondly, the end-to-end performance of the network is of great importance in the study ofcognitive network, so we should establish a low-load and high-efficiency end-to-end monitoringsystem. Based on the index system of cognitive network situation, this paper puts forward a methodfor situation assessment based on improved BP neural network. For the convergence speed oftraditional BP algorithm is slow, and it is easy to fail into local optimal solution, we use the methodof adaptive step size and the simulated annealing algorithm for improvement on BP learningprocedure. Experiments use the assessment method to monitor the specified end-to-end link, andcalculate the corresponding situation level.Thirdly, a complete process of network situational awareness includes understanding the trendsof both current and future network situation. Based on the research of index system of cognitive network situation and end-to-end situation assessment, this paper puts forward a suport vectorregression prediction method that based on particle swarm optimization, and experiment resultshows that the model can meet the requirements in the timeliness and accuracy.