UC San Diego

Wireless sensor networks (WSNs) are envisioned as an approach to a wide variety of monitoring applications. From environmental monitoring to military surveillance, the range of scenarios is diverse, as are the constraints on the networking problem. In most cases, these wireless systems are energy-constrained, and it is impractical or infeasible to replace or recharge the batteries. These systems must be designed with the goal of energy efficiency. In this work, we investigate energy efficient approaches for WSNs that exhibit a variety of connectivity properties. We study three categories of WSNs in particular. In the traditional view of multi-hop sensor networks, there is a large number of sensor nodes with end -to-end connectivity, and they use multi-hop communication. We also consider dense, small-scale sensor networks, in which there are only a few nodes, and communication is primarily single-hop. Lastly, we look at sparse sensor networks, known as a delay- or disruption- tolerant network (DTN). These networks lack end-to-end connectivity, and must rely on mobility to make connections to route data. With these three categories of WSNs, we consider approaches that are specifically applied to each unique scenario. For multi-hop WSNs, we use a distributed approach to Kalman filtering for target tracking. By moving the signal processing onto the local nodes, the RF communication, which typically consumes the most power, is reduced. Through simulation, we evaluate the tracking performance. For our approach to dense, small -scale networks, we exploit the fact that multi-hop communication is rarely needed. We implement a hybrid MAC/ routing protocol that assumes that routing is only needed on rare occasions, thus reducing the idle listening energy consumption. For sparse DTNs, we developed a power management scheme that uses the local information available to determine a node's likelihood of delivery, which is used to decide if a node should sleep or listen/ beacon. We evaluate these two categories of WSNs through simulation and some experimental work (for the small-scale case), and we show significant gains in energy efficiency over existing approaches