Abstract

Traditional path loss (PL) models often fail to capture the spatial characteristics of wireless channels in environments combining line-of-sight (LOS) and non-line-of-sight (NLOS) conditions, where shadow fading exhibits clear spatial non-stationarity. To address this limitation, we propose a Universal Kriging (UK)-based model tailored for predicting shadow fading effects in such scenarios. UK, a method well-suited for non-stationary data, enables more accurate spatial interpolation by incorporating both deterministic trends and stochastic variations. We validate the model using 28 GHz measurements collected in a library environment. Results show that UK consistently outperforms Ordinary Kriging (OK), achieving a 3.7 dB root-mean-square-error (RMSE) reduction in 90% of test cases and reducing the mean RMSE by 69.3% (from 4.85 dB to 1.49 dB). These findings highlight UK's potential for improving PL prediction accuracy in complex LOS/NLOS propagation scenarios.