Authors: Liu, Y.; Wang, S.; Zhang, Y.; Wiart, J.
20th European Conference on Antennas and Propagation (EuCAP), 19 – 24 April, 2026, Dublin, Ireland,
doi: 10.23919/EuCAP68105.2026.11612300
Abstract
The deployment of 5G and beyond networks has heightened public concerns regarding radio frequency electromagnetic field (RF-EMF) exposure. Measurement based assessments are often affected by uncertainties arising from device noise, spatial environmental variability, and temporal fluctuations in base station activity. This paper presents a comprehensive framework on spatial and temporal prediction of RF-EMF exposure with quantified uncertainty. Electric (E) field measurements were collected over 11 months from 12 sensors deployed in an urban environment. On the spatial scale, Polynomial Chaos Expansion (PCE) combined with Bootstrap resampling is used to model and predict exposure levels and spatial uncertainty. On the temporal scale, a Long Short-Term Memory (LSTM) network is used for multi-step forecasting, and conformal prediction is employed to provide statistically valid confidence intervals. The proposed framework demonstrates improved reliability in exposure prediction and uncertainty estimation for continuous EMF monitoring in complex real-world environments.