The performance of in-building measurement-based path loss modelling using kriging

Abstract

An accuracy evaluation analysis of a novel in-building measurement-based path loss prediction narrowband model is presented here, comparing the performance of Kriging-aided shadowing prediction against the most traditional assumption of slow fading as a random variable and a classical estimation derived from linear interpolation. Extensive radio measurements were employed using distinct samples to calibrate (tuning dataset) and validate (testing dataset) the model. Path loss predictions are made over the testing dataset locations to compare it against the measured values, thus obtaining an error in the prediction from the difference between measurements and predictions. The results in the seven buildings evaluated show the potential of Kriging-aided channel modelling with a higher level of confidence than other modelling approaches compared hereafter.