Deep Learning-Based Uplink Transmit Power Estimation in 4G/5G Networks

In urban environments, regulation for radiofrequency electromagnetic field (RF-EMF) exposure limits the intensity of these fields in wireless communication links between a user (UE) and a base station (BS), for health purposes. Therefore, it is relevant to quantify the power emitted by a user over space in time. A comprehensive measurement and assessment of E-field levels by BSs in micro-environments in France has been done by [1]. To be able to predict such levels using publicly available data, the work [2] proposes a convolutional neural network (CNN) method using heatmaps from the physical environment and the radiation patterns of antennas. The present work proposes to combine heatmap images with measurements of downlink (DL) frequencies of BSs to estimate the uplink (UL) power transmit by a UE in a 4G/5G network, using a deep learning framework.