Title: Deep learning techniques for noise-resilient localisation in wireless sensor networks

Authors: Nuha A.S. Alwan; Zahir M. Hussain

Addresses: College of Engineering, University of Baghdad, Baghdad 10011, Iraq ' School of Engineering, Edith Cowan University, Joondalup, WA 6027, Australia

Abstract: Deep learning techniques are highly attractive for extracting meaningful representations of data of the type encountered in wireless sensor network applications. In particular, these techniques lend themselves readily to the implementation of node localisation in wireless sensor networks as the present work demonstrates. Anchor-based range-based time-of-arrival-measurement deep learning localisation techniques are implemented both in the classification and in the regression modes using convolutional neural networks and standard deep neural networks. A performance comparison against gradient descent localisation is carried out. Deep learning techniques exhibited better noise resilience for typical standard deviation ranges of the measurement noise, especially when used in the classification mode. Compressed sampling is also implemented for tracking a mobile node via deep learning regression techniques to increase energy efficiency.

Keywords: deep learning; classification; regression; neural network; WSN; wireless sensor network; localisation.

DOI: 10.1504/IJSNET.2021.115911

International Journal of Sensor Networks, 2021 Vol.36 No.2, pp.59 - 67

Received: 18 Oct 2020
Accepted: 13 Dec 2020

Published online: 23 Jun 2021 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article