Throughput Prediction in Wireless Networks Using Statistical Learning
The focus of this paper is on the estimation of throughput in wireless networks, more specifically on IEEE 802.11. The authors' proposal is based on active measurements and statistical learning tools. They present a methodology where the system is trained during short periods with application flows and probe packets bursts. They learn the relation between throughput obtained by the application and the state of the network, which is inferred from the interarrival times of the probe packets bursts. As a result they obtain a continuous non intrusive methodology that allows determining the maximum throughput of a wireless connection only knowing some characteristics of the network.