Fingerprinting Localization Using Ultra-Wideband and Neural Networks

In this paper, the interest in localization is basically for security issues in a big industry that is still considered to be one of the most dangerous working places, the mines. Mines conditions make solutions based on TOA (time of arrival), AOA (angle of arrival), or even RSS (received signal strength) subject to big errors. The system we discuss relies on fingerprinting technique to overcome those inconveniences. The use of UWB with its high temporal resolution and its multipath beneficial properties will constitute the basic pillar in overcoming much of indoor localization problems. Neural networks with their interpolation characteristics have the role of replacing any database correlator (or search engine) that usually exists in fingerprinting techniques. Measurements of the channel response in the investigated medium, in addition to the choice of the appropriate parameters, will be explored. Evaluation of the neural network performance and comparison with other indoor techniques will help identifying the utility of the proposed system. Finally, future work to be done will be described.