Enhanced Gas Source Localization Using Distributed IoT Sensors and Bayesian Inference

TL;DR

This paper presents a novel approach combining distributed IoT sensors with Bayesian inference to accurately localize gas sources in turbulent environments, demonstrating effectiveness through real-world experiments and simulations.

Contribution

It introduces a Bayesian inference-based algorithm for gas source localization using distributed IoT sensors, validated with both synthetic and real water vapor data.

Findings

Localization error below inter-sensor distance

Effective in turbulent and real-world conditions

Robust performance with synthetic and real data

Abstract

Identifying a gas source in turbulent environments presents a significant challenge for critical applications such as environmental monitoring and emergency response. This issue is addressed through an approach that combines distributed IoT smart sensors with an algorithm based on Bayesian inference and Monte Carlo sampling techniques. Employing a probabilistic model of the environment, such an algorithm interprets the gas readings obtained from an array of static sensors to estimate the location of the source. The performance of our methodology is evaluated by its ability to estimate the source's location within a given time frame. To test the robustness and practical applications of the methods under real-world conditions, we deployed an advanced distributed sensors network to gather water vapor data from a controlled source. The proposed methodology performs well when using both the synthetic data generated by the model of the environment and those measured in the real experiment, with the source localization error consistently lower than the distance between one sensor and the next in the array.