Aggregation in sensor networks: an energy–accuracy trade-off

Abstract

Wireless ad hoc sensor networks (WASNs) are in need of the study of useful applications that will help the researchers view them as distributed physically coupled systems, a collective that estimates the physical environment, and not just energy-limited ad hoc networks. We develop this perspective using a large and interesting class of WASN applications called aggregation applications. In particular, we consider the challenging periodic aggregation problem where the WASN provides the user with periodic estimates of the environment, as opposed to simpler and previously studied snapshot aggregation problems. In periodic aggregation our approach allows the spatial–temporal correlation among values sensed at the various nodes to be exploited towards energy-efficient estimation of the aggregated value of interest. Our approach also creates a system level energy vs. accuracy knob whereby the more the estimation error that the user can tolerate, the less is the energy consumed. We present a distributed estimation algorithm that can be applied to explore the energy–accuracy subspace for a subclass of periodic aggregation problems, and present extensive simulation results that validate our approach. The resulting algorithm, apart from being more flexible in the energy–accuracy subspace and more robust, can also bring considerable energy savings for a typical accuracy requirement (fivefold decrease in energy consumption for 5% estimation error) compared to repeated snapshot aggregations.