1.
Prabhakar
– Malla Reddy Engineering College for Women, Hyderabad, Telangana, India.
2.
Vamsi Krishna M
– Malla Reddy Engineering College for Women, Hyderabad, Telangana, India.
3.
Y. Geetha Reddy
– Malla Reddy Engineering College for Women, Hyderabad, Telangana, India.
Abstract
Urban air pollution is a major environmental and public health issue. It needs accurate measurements and timely forecasts to help with decision-making. This study introduces an integrated framework for high-resolution air quality monitoring and prediction (AQMP). It combines mobile sensing, edge intelligence, and hybrid modeling. To get around the limits of fixed monitoring stations, we use unmanned aerial vehicles (UAVs) with lightweight sensors and 5G modules. These collect real-time data on pollutants like PM2.5, PM10, CO, CO2, and O3 at different altitudes and locations. To deal with the issues of low-cost hardware, we apply TinyML directly on microcontrollers. We use denoising autoencoders for local missing data and convolutional neural networks for efficient parameter prediction without needing cloud services. We improve ata reliability with a hybrid design that applies Empirical Mode Decomposition (EMD) to smooth non-stationary data into sub-series and uses truncated Singular Value Decomposition (SVD) to reduce noise and find hidden relationships between nearby stations and various pollutants. To facilitate low-latency communication, the framework utilizes a 5G-enabled message queueing telemetry transport (MQTT) protocol to transmit sensor data to a centralized web server for real-time visualization.
Keywords Air quality prediction, Convolutional Neural Networks (CNN), Edge intelligence, Empirical Mode Decomposition (EMD), Hyper-heuristic model
