On the Cross-Application of Calibrated Pathloss Models using Area Features: Finding a Way to Determine Similarity Between Areas

Published

February 1, 2020

Author(s)

Jiayi Zhang, Camillo Gentile, Wesley Garey

Abstract

Pathloss-model calibration is the practice of refining the nominal parameters of a model according to measurement samples collected in a specific area. It is a widely used by mobile providers to reduce prediction error up to tens of dBs depending on the model category. It comes, however, at the expense of both time and monetary resources. Because the resources required to calibrate all deployment areas are prohibitive, a selection from a representative set of calibrated models can be applied to an are with no measurement data; we refer to this practice as model cross-application. How well the model predicts will depend on the similarity between the two areas. In this article, we propose a methodology for cross-application in which we identify the most effective features to determine area similarity. To do so, we analyzed over three million measurement samples from five metropolitan regions throughout the United States - comprising urban, suburban, and rural environments - while considering a broad range of model categories, from purely empirical to highly deterministic. We also validated the performance of the models per environment, both in terms of absolute prediction error and in terms of error reduction due to calibration.

Citation

IEEE Antennas and Propagation Magazine

Volume

62

Issue

1

Pub Type

Journals

Download Paper

https://doi.org/10.1109/MAP.2019.2943272

Local Download

Keywords

prediction, tuning, signature, fingerprint

Citation

Zhang, J. , Gentile, C. and Garey, W. (2020), On the Cross-Application of Calibrated Pathloss Models using Area Features: Finding a Way to Determine Similarity Between Areas, IEEE Antennas and Propagation Magazine, [online], https://doi.org/10.1109/MAP.2019.2943272, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=926412 (Accessed April 23, 2024)