Data-driven propagation modeling for a class of IEEE 802.15.4 wireless devices in an indoor environment

# 1


We consider the problem of deployment of static indoor wireless networks for connecting sensors to a data collection station, in the context of Internet of Things (IoT) applications. The deployment of such indoor wireless networks requires the ability to predict the quality of the wireless link between any desired pair of points in the indoor environment. Modelling the channel propagation environment in factories and buildings is, however, challenging. We adopt the methodology of spatial sampling of a large number of links, exchanging packets between actual devices placed at the ends of these links, and then using the collected RSSI (received signal strength) data to develop a predictive link model. We highlight three issues in such a data-driven modelling approach. First, the limited range of link lengths over which data are collected can affect the accuracy of the estimated channel parameters. Second, due to device characteristics, packet error rate (PER) variation with an average signal to noise ratio (SNR) may be significantly different from that predicted by theory. Third, RSSI estimates based on successfully received packets suffer from success bias. Our proposed methodology overcomes these three issues via targeted sampling of link lengths, characterization of PER versus RSSI via controlled measurements on the transceiver devices, and an EM-like (expectation-maximization) framework to handle lost packets via suitable imputation of RSSI on lost packets. We validate our methodology on a generative model and then test it on data collected from field experiments, to quantify the gains coming from the EM framework. Even though our indoor environment with over three hundred links has intercepting walls of different types and numbers, doors and windows of different sizes, a linear path-loss model, superposed with a suitable Nakagami-m fading model atop log-normal shadowing, provides a good fit to the experimental data. By using ten-fold cross validation over our sample of over 500 links, we also report on the efficacy of our model in predicting the packet error rates on links.

Anitha Varghese

Anitha Varghese is working as a senior scientist in ABB corporate research, Bangalore. She is currently pursuing her Ph.D in ECE Department, IISc, Bangalore under the guidance of Prof. Anurag Kumar. She received her ME in Telecommunications engineering from the ECE Department, IISc in 2006, and her B.Tech in electronics and communication engineering from Kerala University in 2004. She worked as a researcher in General Motors India Science lab, from 2006-2010. Her research interests include design and analysis of wireless communication networks, and communication security in the context of industrial automation.