Abstract

Air pollution poses a growing threat to public health, urban sustainability, and environmental balance. Traditional air monitoring stations, though accurate, are expensive and offer limited spatial coverage. The combination of wireless sensor networks (WSNs) with intelligent predictive models provides a low-cost and scalable solution for real-time air quality assessment. Field Programmable Gate Arrays (FPGAs) further enhance system performance by offering high-speed data processing and energy-efficient computation at the edge. Despite technological progress, existing air quality monitoring frameworks often suffer from high latency, limited adaptability to dynamic conditions, and excessive power consumption. In addition, the fluctuating nature of environmental parameters such as temperature, humidity, and particulate matter (PM2.5, PM10) necessitates a model capable of learning temporal dependencies for accurate forecasting. This work presents an FPGA-induced WSN architecture integrated with Long Short-Term Memory (LSTM) networks for predictive air pollution monitoring. The proposed system deploys distributed sensor nodes equipped with low-power microcontrollers and FPGA modules for edge data pre-processing. Sensor data streams are transmitted via wireless nodes to a centralized unit running an optimized LSTM model for pollutant prediction. The FPGA accelerates matrix computations, reducing inference latency, while adaptive data sampling minimizes energy usage. The model is trained and tested on real-time datasets containing concentrations of CO2, NO2, and PM2.5 from urban monitoring sites. Experimental evaluation demonstrated that the proposed FPGA-enabled LSTM system achieved a Root Mean Squared Error (RMSE) of 3.5–3.6, a Mean Absolute Error (MAE) of 2.8–2.9, and a Mean Absolute Percentage Error (MAPE) of 1.5–1.6% over 10 sensor nodes. Energy consumption per node was reduced to 9.5–9.6 J, while prediction latency was lowered to approximately 95 ms, outperforming traditional ANN, regression, and CPU-based LSTM methods. The framework exhibited high scalability and real-time predictive capability, confirming its effectiveness for low-latency, energy-efficient, and accurate urban air quality monitoring.

Authors

Machhindranath M. Dhane, M. Radhika
Government First Grade College, Kolar, India

Keywords

FPGA, Wireless Sensor Networks, LSTM, Air Pollution Prediction, Edge Computing