Field Reconstruction in Sensor Networks with Coverage Holes and Packet Losses

TL;DR

This paper analyzes the accuracy of field reconstruction in wireless sensor networks with coverage gaps and packet losses, using asymptotic spectral analysis to evaluate the mean square error of linear filtering methods.

Contribution

It provides a novel asymptotic analysis of the spectral properties of the sampling matrix in sensor networks with coverage holes and packet losses, enabling performance evaluation.

Findings

Derived moments and spectral distribution of VV* as network size grows

Provided an approximation for mean square error of field estimation

Analyzed the impact of coverage holes and packet losses on reconstruction quality

Abstract

Environmental monitoring is often performed through a wireless sensor network, whose nodes are randomly deployed over the geographical region of interest. Sensors sample a physical phenomenon (the so-called field) and send their measurements to a {\em sink} node, which is in charge of reconstructing the field from such irregular samples. In this work, we focus on scenarios of practical interest where the sensor deployment is unfeasible in certain areas of the geographical region, e.g., due to terrain asperities, and the delivery of sensor measurements to the sink may fail due to fading or to transmission collisions among sensors simultaneously accessing the wireless medium. Under these conditions, we carry out an asymptotic analysis and evaluate the quality of the estimation of a d-dimensional field when the sink uses linear filtering as a reconstruction technique. Specifically, given the matrix representing the sampling system, V, we derive both the moments and an expression of the limiting spectral distribution of VV*, as the size of V goes to infinity and its aspect ratio has a finite limit bounded away from zero. By using such asymptotic results, we approximate the mean square error on the estimated field through the eta-transform of VV*, and derive the sensor network performance under the conditions described above.