Neural Networks for Fingerprinting-Based Indoor Localization Using Ultra-Wideband

This paper discusses the use of neural networks in an underground radio-localization system. In a highly aggressive environment such as mines, reliability and robustness are essential to any operational system. Using UWB as the physical wireless propagation medium and combined with fingerprinting-geolocation and neural networks, this work tends to overcome many of the problems encountered in indoor environments. Full description of the system and the adopted approach will help accentuate the role of neural networks in improving the overall performance. Moreover a comparison between MLP and RBF performance is presented, providing a clear evidence of the role and importance of the neural networks in offering good accuracy and precision to the final system. Index Terms Indoor localization, LOS, NLOS, fingerprinting-localization, ultra-wide-band (UWB), impulse response (IR), neural networks (NN), multi-layer perceptron (MLP), radial basis functions (RBF).