Localization for anchoritic sensor networks

Abstract

We introduce a class of anchoritic sensor networks, where communications between sensor nodes are undesirable or infeasible due to, e.g., harsh environments, energy constraints, or security considerations. Instead, we assume that the sensors buffer the measurements over the lifetime and report them directly to a sink without necessarily requiring communications. Upon retrieval of the reports, all sensor data measurements will be available to a central entity for post processing. Our algorithm is based on the further assumption that some of the data fields that are being observed by the sensors can be modeled as a local (i.e. having decaying spatial correlations) stochastic process; if not, then choose an auxiliary field, e.g., carefully engineered random signals intentionally generated by arranged devices, "cloud shadows" cast on the ground, or animal heat. The sensor nodes record the measurements, or a function of the measurements, e.g., "1" when the measured signal is above a threshold, and "0" otherwise. These time-stamped sequences are ultimately transferred to the sink. The localization problem is then approached by analyzing the correlations between these sequences at pairs of nodes. As for applications, we discuss the localization scheme for large-scaled sensor networks deployed on the seabed and study a two-tiered architecture that organizes deaf sensors with local masters. © Springer-Verlag Berlin Heidelberg 2007.