Nonlinear Network Od Flow Estimation

The origin–destination (OD) traffic matrix in local area networks is useful for solving problems in network design, routing, configuration debugging, and monitoring. However, with the rapid development of the Internet, the structure of the network has changed thoroughly. It is hard or expensive to measure it direct. Network tomography as a method of end-to-end through indirect measurement data to infer the state of network parameters on the technology is becoming a hot oneBecause of the link counts are usually far less than the OD counts, OD Traffic Matrix Estimation is a typical dubious problem. Usually, we often assume that each OD shloud obey a model (such as the Gaussian distribution, etc.) or the same time a number of OD counts obey a model (such as gravity model, etc.). It is believe that OD flows change is a complex process, and it is difficult to use a certern model to solve the problem. However, there is a certain relevance between each traffic matrix in the same network.We can take advantage of some rules from the history of the origin–destination traffic matrix. We do some study in the following two aspects.(1) We believe that OD flows are rich to fully capture spatial and temporal correlations.We build the recurrent multilayer perceptron (RMLP) network to estimate the origin–destination traffic matrix. We propose a new method to estimate the origin–destination traffic matrix which bases on the recurrent multilayer perceptron networks. We apply the RMLP model to our nonlinear and dynamic system that can be used for both estimation and prediction of the origin–destination (OD) traffic matrix.We make use of link data and some OD flows to get the relationship between link data and OD flows.And then we can predict OD flows with the relationship. Under the implementation of the different constrains and incorporated with partical estimate measurement, we get appealing result.The result shows that the algorithm greatly improves OD flow matrix of the estimated accuracy.(2) We novel the algorithm to estimate the origin–destination traffic matrix which bases on simulated annealing algorithm. Obviously, this model is nonlinear and dynamic system.We take advantage of the relationship between different times of the origin–destination traffic matrix.We can get a better result by increasing the state transition matrix.The results show that our algorithm is validity.