AI-Driven Electromagnetic Radiation Monitoring System for Sustainable Development Goals

Abstract

The large-scale deployment of 5G telecommunications networks, Internet of Things infrastructure and smart cities have greatly increased the public’s exposure to electromagnetic radiation (EMR) in cities. Traditional point-in-time methods of monitoring, however, cannot accurately measure EMR due to the rapidly evolving and heterogeneous environments associated with today’s EMR landscape. This study proposes a complete artificial intelligence (AI)-based system comprising a 22-node Internet of Things (IOT) sensor network, a hybrid convolutional neural network (CNN)-long short-term memory (LSTM) deep learning classifier, mathematically validated models of EMR exposure, and a cloud-based real-time sustainable development goals (SDG) compliance dashboard. Using this system, the CNN-LSTM classifier demonstrated a classification accuracy level of 94.6% for samples collected from the four ICNIRP defined EMR exposure zones, resulting in improved performance in comparison to the basic performance of CNN (87.2%), LSTM (85.9%), random forest (83.5%), and AdaBoost (81.8%). For the four parameters measured, including the power density (S), specific absorption rate (SAR), free-space path loss (FSPL), and exposure index (EI), the continuous values produced by the system matched experimentally validated values with a mean absolute error (MAE) of less than 3.8%. Using Hyderabad as a test city between August 2024 and January 2025, the system had a detection lag time of 1.8 seconds, a false-positive rate of 6.2% and a precision rate of 91.5 % for the validation of 47 cloud-generated alerts to support the achievement of sustainable development goals.