Capacity- and energy-aware activation of sensor nodes for area phenomenon reproduction using wireless network transport

Abstract

We consider a sensor network involving sensors placed in specific locations. An area phenomenon is detected and tracked by activated sensors. The area phenomenon is modeled to consist of K spatially distributed point phenomena. The activated sensors collect data samples characterizing the parameters of the involved point phenomena. They compress observed data readings and transport them to a processing center. The center processes the received data to derive estimates of the point phenomena's parameters. Our sensing stochastic process models account for distance-dependent observation noise perturbations as well as noise correlations. At the processing center, sample mean calculations are used to derive the estimates of the underlying area phenomenon's parameters. We develop computationally efficient algorithms to determine the specific set of sensors for activation under capacity and energy resource constraints so that a sufficiently low reproduction distortion level is attained. We derive lower bounds on the realizable levels of the distortion measure. Using illustrative cases, we demonstrate one of our algorithms to yield distortion levels that are very close to the lower bound, while other lower-complexity schemes often yield distortion levels relatively close to the lower bound.