A Distributed Range-Free Localization WSN Deployment Using Machine Learning Techniques for Air Pollution Monitoring

Abstract

Accurate and energy-efficient localization of wireless sensor nodes is crucial for reliable air quality monitoring in cities. Traditional range-based methods such as Time of Arrival (ToA) and Received Signal Strength Indicator (RSSI) are limited by hardware complexity and the need for calibration, making them less suitable for large-scale deployments. While range-free algorithms are computationally simpler, they are prone to cumulative hop-count errors and irregular radio signals, which reduce accuracy. This paper presents a new framework called Distributed Machine Learning Assisted Range-Free Localization (DML-RFL). It merges anchor-assisted distributed localization with ensemble machine learning techniques to achieve highly precise node positioning in urban settings. DML-RFL begins with DV-Hop and Centroid Localization for initial estimates, then refines these by inputting features into models like Random Forest, Support Vector Machine, Gradient Boosting, and Artificial Neural Networks. The framework is structured within a five-layer architecture that includes sensor, communication, edge processing, machine learning inference, and environmental monitoring dashboard layers, each tailored to support diverse pollution sensors such as PM2.5, PM10, CO, CO₂, SO₂, NO₂, and O₃. Extensive simulations across network densities of 100 to 500 nodes and deployment areas from 500×500 m² to 1500×500 m² show that DML-RFL attains a mean localisation error of 0.31 m in optimal setups, extends network lifetime by 37.4% compared to DV-Hop, and achieves over 97.2% packet delivery ratio. Comparative tests against DV-Hop, APIT, Centroid Localisation, and Weighted Centroid Localisation confirm statistically significant improvements in all key metrics. The proposed system is scalable, cost-effective, and ready for deployment in next-generation urban air-pollution monitoring systems.