Multisensor Data Fusion for Water Quality Monitoring using Wireless Sensor Networks

TL;DR

This paper explores hierarchical wireless sensor networks for water quality monitoring, proposing advanced fusion algorithms at local centers that outperform traditional methods, especially under increased sensor count or contamination.

Contribution

It introduces novel hard- and soft-decision MAP algorithms for decision fusion at local centers in water quality sensor networks, enhancing detection accuracy and reducing costs.

Findings

Proposed algorithms outperform conventional fusion techniques.

Performance gap widens with more sensors or contamination.

Lower total cost in detection accuracy.

Abstract

In this paper, the application of hierarchical wireless sensor networks in water quality monitoring is investigated. Adopting a hierarchical structure, the set of sensors is divided into multiple clusters where the value of the sensing parameter is almost constant in each cluster. The members of each cluster transmit their sensing information to the local fusion center (LFC) of their corresponding cluster, where using some fusion rule, the received information is combined, and then possibly sent to a higher-level central fusion center (CFC). A two-phase processing scheme is also envisioned, in which the first phase is dedicated to detection in the LFC, and the second phase is dedicated to estimation in both the LFC and the CFC. The focus of the present paper is on the problem of decision fusion at the LFC: we propose hard- and soft-decision maximum a posteriori (MAP) algorithms, which exhibit flexibility in minimizing the total cost imposed by incorrect detections in the first phase. The proposed algorithms are simulated and compared with conventional fusion techniques. It is shown that the proposed techniques result in lower cost. Furthermore, when the number of sensors or the amount of contamination increases, the performance gap between the proposed algorithms and the existing methods also widens.