Opportunistic casting for wireless sensor networks

Abstract

Sensor networks of the future are predicted to be densely populated and consisting of tiny sensor nodes that are located mere milimetres from the surface of the ground or other environments. To best utilise these networks, information about the propagation channel is needed along with ways to exploit the channel and network characteristics. This thesis is divided into two main portions - the first is the measurement and characterisation of narrowband RF propagation for surface-level wireless communications; the second is the exploitation of node density to improve communication performance and energy efficiency in wireless sensor networks. The first part of this dissertation reports the results of a measurement campaign for outdoor, ground-surface level wireless communications that stretches over a period of more than 6 months. A comprehensive study of ground level propagation is made for the large-scale path loss, small-scale fading, and time-varying small-scale fading due to movement in the environment. Results for the large-scale path loss show that by incorporating the Norton surface wave component into the signal propagation, a very accurate predictor of the average signal strength can be obtained based on the antenna height and ground properties. Static small-scale fading distributions were also shown to be Rician distributed with a K-factor that depends on the distance from the signal source. Based on the measurement results from irregular terrain, a surface-level irregular terrain (SLIT) model is presented to account to irregular terrain shapes for more accurate path loss prediction. Results for time-varying fading show that the fading caused by wind-blown foliage is Nakagami-$ \emph{m}$ distributed, with a $\emph{m}$ shape factor that is dependent on the wind speed, excess path loss, and seasonal changes to the foliage. Fading due to human movement is difficult to classify for a single person. However, fading for groups of humans moving about can be...