New methods for estimating/forecasting link bandwidths in 802.11b WLANs

Accurate assessment of the available link bandwidths in wireless networks is crucial for applications to negotiate, adapt and control quality-of-service parameters. We focus on new methods for estimating/predicting saturated throughputs (link bandwidths) in 801.11b wireless local area networks (WLANs). We propose a principle component analysis (PCA) and clustering approach for the estimation problem and a pattern modeling and recognition system (PMRS) and method of moments matching (MM) for the prediction problem. A trade off between the complexity and the accuracy of the estimation/prediction algorithms is discussed and compared with a neural networks (NN) approach.