A Novel Air Quality Assessment Method With Low-Cost Sensors and Predictive Analytics Using Autonomous Drones

ISBN: 979-8-89480-841-3


This project presents a novel method for real-time air quality monitoring using a low-cost, modular sensor system designed for future drone integration. Traditional monitoring stations are expensive, stationary, and lack spatial resolution, while most portable systems offer limited accuracy or predictive capability. This work bridges that gap by combining affordability, mobility, and intelligent analytics into a single, scalable solution.

The prototype uses a Raspberry Pi platform integrated with MQ-series gas sensors (CO, NO℃, NH℃) and a BME680 environmental sensor. Data is collected at 15-second intervals, timestamped using NTP synchronization to minimize drift, and processed through machine learning models. A Random Forest Regressor was trained on the AirQualityUCI dataset to predict pollutant levels, while DBSCAN was used to identify pollution anomalies. All models were optimized for on-device inference using TensorFlow Lite with quantization.

The system was tested in three distinct environments urban intersections, residential neighborhoods, and construction zones over six 30-minute sessions. Performance metrics showed strong predictive accuracy with R² scores of 0.85 for CO, 0.81 for NO℃, and 0.91 for NOₓ. Mean Absolute Errors ranged from 0.071 to 0.156 ppm, with real-time prediction latency of ~1 second and average uptime of 5.5 hours per battery cycle. Compared to consumer-grade and government-grade monitoring systems, the proposed model offers a competitive balance of cost, accuracy, and real-time capability.

This approach lays the foundation for future drone-based deployments, enabling rapid, distributed air quality surveillance in both urban and underserved regions

References

  • Y. Feng, X. Long, "Comparative Analysis of Traditional Sensors and Machine Learning Enhanced Sensors," 2024.
  • IBM, "What is random forest?" 2024. Retrieved from https://www.ibm.com/think/topics/random forest
  • A. Rahman et al., "Enhancing IoT-Based Environmental Monitoring and Power Forecasting," 2024.
  • ScienceDirect, "Unmanned Aerial Vehicle (UAV)," 2024. Retrieved from https://www.sciencedirect.com/topics/agricultural-and-biological-sciences/unmanned-aerial-vehicle
  • UCI, "Air Quality Dataset," 2024. Retrieved from https://archive.ics.uci.edu/dataset/360/air+quality
  • R. Van Dingenen, F. J. Dentener, F. Raes et al., "The global impact of ozone on agricultural crop yields under current and future air quality legislation," Atmospheric Environment, vol. 43, no. 3, pp. 604 618, 2009.
  • World Bank, The Global Burden of Air Pollution: Impacts on Health and the Economy, 2022.
  • World Health Organization, "Ambient (outdoor) air pollution," 2024. Retrieved from https://www.who.int/news-room/fact-sheets/detail/ambient-(outdoor)-air-quality-and-health
  • L. Zhou et al., "Comparative Analysis of Linear Regression, Neural Networks, and Random Forest Regression for Predicting Ozone," 2023.
  • R. Chen et al., "Urban Air Quality Mapping via Campus Transit Sensors: A Case Study," 2022.
  • U.S. Environmental Protection Agency, Air Sensor Toolbox for Citizen Scientists, Researchers and Developers, 2021.
  • F. Parra-Medina et al., "Low-Cost Solution for Air Quality Monitoring: UAV and LoRa Protocol," 2024.