Deterministic Ray Tracing (RT) is widely used for coverage prediction and network planning in 5G New Radio systems. Jointly with reliable environmental reconstruction, RT accuracy depends strongly on antenna radiation pattern modeling, especially at millimeter-wave (mmWave) frequencies. This paper compares three antenna modeling strategies implemented in the Matlab 5G Toolbox and validates their predictions against an extensive 27 GHz urban walk-test dataset. The assessment combines point-wise error metrics, empirical cumulative distribution functions, and spatial correlation analysis. Results highlight the absence of an optimal antenna model; rather, the provided accuracy is context-specific. The beam-based model achieves the best agreement in guided propagation conditions, whereas the envelope and isotropic models yield more robust macroscopic coverage estimates in heavily obstructed areas affected by incomplete environmental representation and unmodeled clutter.
Salvo et al. (Wed,) studied this question.