Estimation of available bandwidth share by tracking unknown cross-traffic with adaptive extended Kalman filter

We propose a nonlinear network model to estimate the available bandwidth share of a source by tracking the unknown cross-traffic. Especially sudden changes in cross-traffic behavior are challenging to adapt since measurement or process models of the existing algorithms generally do not include the cross-traffic in the model. As a novel approach, combined cross-traffic behavior, generally considered as additive noise, is modeled as an unknown source enabling tracking of both the cross-traffic and network behavior. Adaptive Extended Kalman Filter with Unknown Inputs (EKF-UI) is used for the estimation of available bandwidth share. This approach works recursively and is suitable for real-time applications. Moreover, the measurements are based on passive monitoring. Hence, no probe traffic is induced to the network. It is also shown with multiple simulations that this model is robust against variable network conditions.