Measurement and prediction of path loss at 153.87 MHz to support mission-critical public safety networks in Minas Gerais State

Abstract

This dissertation investigates path loss prediction for mission-critical public-safety networks operating in the Very High Frequency (VHF) band under heterogeneous propagation conditions. In this sense, a georeferenced measurement campaign was conducted using the operational Digital Mobile Radio (DMR) network of the Minas Gerais Military Fire Brigade (CBMMG) in Juiz de Fora, Brazil, where irregular terrain, vegetation, buildings, water bodies, and transitions among urban, suburban, and rural environments affect radio propagation. The received-signal-strength data were processed to obtain local-mean path loss values and evaluate empirical, regression, and machinelearning (ML) models. The empirical models were assessed using standard and locally calibrated parameterizations, while the regression and ML models used geometric, topographic, and georeferenced features. The empirical models showed limited ability to represent the measured propagation behavior, even after calibration. The regression models improved prediction accuracy, particularly the Polynomial Regression model. The ML models achieved the best overall performance, with the Random Forest model providing the highest accuracy and strongest generalization across both environmentspecific and combined datasets. These findings support using data-driven ML models for path loss prediction and radio planning in mission-critical VHF public-safety networks.