Dependable Dempster-Shafer Inference Framework for Urban Air Quality Monitoring

Observations of air pollution have become an increasing concern for authorities and citizens due to emissions from population growth and urbanization. In this regard, low-cost wireless sensor networks have emerged as a popular, cost-effective solution for monitoring and estimation of air pollutant levels in local areas. Due to environmental vulnerability, ensuring the required performance and reliability of these sensing devices remains an open problem. This paper presents the development of a dependable Dempster-Shafer inference (DDSI) framework based on evidence theory for colocated low-cost sensors for air quality monitoring. This approach facilitates fault detection and accounts for uncertainty to enhance the performance of the sensor network. A switching mechanism is employed in the inference layer to select the most reliable information based on the probability of the current operational status of sensors, thereby leveraging the resilience and data integrity of the monitoring system. The proposed method is comprehensively validated in both laboratory and real-world settings. The DDSI framework is then tested in monitoring humidity, temperature and particulate matters in the suburbs. The results are benchmarked with data collected from state-run air quality monitoring stations. Statistical analysis is conducted to show the accuracy of the proposed framework and a high alignment with station observations, achieving an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of up to 0.918 for meteorological parameters and 0.892 for fine particles (PM<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2.5</sub>).