Handbook on Array Processing and Sensor Networks

CHAPTER 17

Distributed Algorithms in Sensor Networks

Usman A. Khan, Soummya Kar, and José M. F. Moura

Carnegie Mellon University, Department of Electrical and Computer Engineering, Pittsburgh, Pennsylvania

17.1 INTRODUCTION

Advances in integrated electronics, radio-frequency (RF) technologies and sensing devices have made it possible to deploy large numbers of cheap sensors for the purposes of monitoring, tracking, estimation, and control of complex large-scale dynamical systems through collaborative signal processing [1–3]. For example, consider a detection problem where the state of the environment is monitored “locally” by sensors; each sensor makes a measurement, based on which it may make a local decision—the current state of the sensor. A problem of interest is how to fuse these local decisions. An approach is to send these states to a fusion center where the optimal detector is formulated; this has been considered extensively in the literature since the early work in [4–6], the book by Varshney [7], and, more recently [8, 9]. This centralized or parallel architecture, which may have several advantages, is neither robust nor scalable when the size of the sensor network grows because of resource (bandwidth, power, etc.) constraints at the sensors and because it has a single point of failure. An alternative architecture for resource-constrained networks is a weblike topology ...