A power-efficient IoT mechanism with adaptive recurrent temporal optimized convolutional learning (ARTOCL) scheme for real-time urban air quality monitoring

This paper proposes a power-efficient internet of things (IoT) mechanism with an adaptive recurrent temporal optimized convolutional learning (ARTOCL) scheme for real-time urban air quality monitoring. The proposed system integrates energy-aware IoT nodes and an IoT gateway to create a peer-to-peer network. IoT sensors collect PM2.5 data from different regions and transmit it to an IoT server for real-time tracking, control, and logging. The solution is designed to maximize energy savings through duty cycling while implementing the ARTOCL scheme, which combines long short-term memory (LSTM), Autoregressive Integrated Moving Average (ARIMA), and convolutional neural networks (CNN). The PM2.5 data is accessible through an Android app for visualization. A continuous air quality data collection experiment was conducted in an outdoor area to validate the system’s functionality. PM2.5 concentrations were analyzed to evaluate air quality across different locations. The results showed significant improvements in error metrics for PM2.5, with reductions in MAE, RMSE, and nRMSE to 0.3, 0.2, and 0.060, respectively. The accuracy metrics also improved, with the R² value for PM2.5 rising to 0.7 when using the ARTOCL algorithm. Additionally, the energy consumption analysis confirms the robustness and impressive power-saving efficiency of the proposed IoT solution.