Energy efficiency and spatiotemporal sampling in wireless sensor networks

Seema Bandyopadhyay, Purdue University

Abstract

Recent advances in wireless communications and MEMS have motivated the development of extremely small and low-cost sensors. Multi-hop, ad-hoc wireless networks of large numbers of such sensors can be effective tools for gathering data in a wide variety of environments. A key determinant of the effectiveness of these networks is their longevity, which is limited by the energy resources of each sensor. Hence, in the first part of this thesis, we (i) analytically determine the energy cost of communicating data from all sensors to a central node, and (ii) minimize this cost via optimal selection of the parameters in a hierarchical, distributed clustering algorithm. This clustering algorithm is found to be significantly more energy efficient than other existing algorithms. Another critical determinant of the effectiveness of a sensor network is its spatiotemporal sampling strategy. In the second part of the thesis, we address this issue and obtain results that include: (i) characterization of the trade-offs among sensor density, energy usage, temporal sampling rates and spatial sampling rates in wireless sensor networks, and (ii) derivation of a lower bound on the delay in gathering packets at a given spatial sampling rate. These results are obtained for an idealized media-access scheme, but they can serve as a benchmark to evaluate the performance of other protocols. In the third part of the thesis, we consider a wireless network of mobile sensors and characterize the time to detect any event by at least one of these sensors. In the course of characterizing the event detection time, we obtain closed form solutions in the time-domain for the transient and steady state probabilities of correlated random walks on a finite lattice in one and two dimensions. We also obtain closed form solutions for the Laplace transforms of the first passage time for such walks. These results allow one to determine the required density of sensors in a network to detect events with high probability in a given time period. Moreover, the results related to correlated random walks can be applied to many other applications in physical and chemical sciences.

Degree

Ph.D.

Advisors

Coyle, Purdue University.

Subject Area

Electrical engineering|Computer science

COinS