Estimating resources in wireless ad-hoc networks: A Kalman filter approach

The sparse and fluctuating bandwidth that characterize wireless links in ad-hoc networks present challenging conditions in our goal to ensure Quality of Service (QoS) in wireless connections and to maintain a satisfactory Quality of Experience (QoE) for the end user. In this environment, it is not only necessary to determine the current available bandwidth for a given link, but also to predict it for the next transmission period. We present a predictive available bandwidth estimation method, based on a Kalman filter in conjunction with a change detection mechanism. This approach achieves both good suppression of the high-frequency variations and fast tracking of the changes of the available bandwidth estimation.