Robust Detection of Random Events with Spatially Correlated Data in Wireless Sensor Networks via Distributed Compressive Sensing

TL;DR

This paper introduces a robust, nonparametric detection method for random events in dense wireless sensor networks, leveraging compressive sensing to efficiently utilize correlated data without prior noise knowledge.

Contribution

It develops a novel detection approach exploiting covariance matrix sparsity via distributed compressive sensing, enhancing robustness and efficiency in sensor networks.

Findings

Effective detection without noise parameter knowledge

Robust performance with sparse random projections

Utilizes covariance sparsity for improved detection

Abstract

In this paper, we exploit the theory of compressive sensing to perform detection of a random source in a dense sensor network. When the sensors are densely deployed, observations at adjacent sensors are highly correlated while those corresponding to distant sensors are less correlated. Thus, the covariance matrix of the concatenated observation vector of all the sensors at any given time can be sparse where the sparse structure depends on the network topology and the correlation model. Exploiting the sparsity structure of the covariance matrix, we develop a robust nonparametric detector to detect the presence of the random event using a compressed version of the data collected at the distributed nodes. We employ the multiple access channel (MAC) model with distributed random projections for sensors to transmit observations so that a compressed version of the observations is available at the fusion center. Detection is performed by constructing a decision statistic based on the covariance information of uncompressed data which is estimated using compressed data. The proposed approach does not require any knowledge of the noise parameter to set the threshold, and is also robust when the distributed random projection matrices become sparse.