A Neural Method for Identifying Transmission Source Locations

In recent years, there has been great interest in node localization within low-power communication networks. These technologies include Bluetooth, GPS, IEEE 802.11, and other transmission protocols. Most techniques are based on variations in the RF signal-to-noise ratio, but this paper introduces a new method, which employs packet statistics. In this work, packet information was collected from several stationary clients while moving a portable server and access point. Packet statistics and the corresponding server locations were subsequently used to train neural networks. Our studies have shown that the networks can determine the location of additional transmitters based on the packet histories of the stationary clients