中文摘要：

现有网络链路参数估计方法大都假设网络链路状态在测量周期内是平稳的,不能获得网络链路参数的时变特征。该文提出了一种非平稳的网络链路丢包率层析成像方法。假定在一个相对较小的时窗内,丢包率随时间变化的曲线可用一个k阶可导的函数来描述;用网络层析成像的方法求得这些函数的k阶泰勒展开式;然后根据各时窗内的逼近结果,用反比距离加权估计整个测量周期内链路的时变丢包率。NS2仿真验证了该方法能有效追踪链路丢包率的变化,且优于现有的网络链路丢包率层析成像方法。

英文摘要：

Most of network link parameters inference methods assume that link states are stationary during measurement period,and can not obtain time-varying characteristics of link parameters.In this paper,a novel nonstationary internal loss tomography method is proposed.Assume in a relatively short time window,the time-varying curves of link loss rates are described by k times continuous differentiable functions.The k-th order Taylor Serieses of these functions are estimated using network tomography approach.Then based on the estimates of each time window,the time-varying link loss rates of entire measurement period are obtained by integrating the estimates of all time windows using Inverse Distance Square Weighted algorithm.NS2 simulations show that the method is capable of tracking variation of link loss rates effectively,and superior existing stationary internal loss tomography methods.