Sensor Placement for Distributed Sensing

Abstract

Sensor networks are omnipresent in our daily lives. In this chapter, we deal with the problem of organizing such a sensor network, or in other words, how to best place the different sensors. The sensing applications we consider in this chapter yield the estimation and detection of an unknown variable that can be indirectly measured by different sensors. We assume that the number of sensors is limited, e.g., because of cost considerations, and hence the problem boils down to optimally placing the smallest number of sensors such that a certain inference performance can be guaranteed. Different types of inference metrics for both estimation and detection will be studied. A major contribution of this work involves the treatment of not only conditionally independent measurements yet also conditionally dependent ones, which makes the problem far from trivial. Our results will be corroborated by a running example on field estimation and detection.