Machine-Learning and 3D Point-Cloud Based Signal Power Path Loss Model for the Deployment of Wireless Communication Systems

• Published in: IEEE access Vol. 7; pp. 42507 - 42517
• Main Authors: Egi, Yunus, Otero, Carlos E.
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: 3D image segmentation; Algorithms; Artificial neural networks; Canopies; deployment; Empirical analysis; Environment models; Estimation; Feature extraction; Frequency analysis; GPS; Image classification; image fusion; IMU; Laser radar; LiDAR; Machine learning; Optical wavelength conversion; Parameter estimation; path loss; Radio frequency; RSL; Rural areas; Rural environments; Solid modeling; Three dimensional models; Three-dimensional displays; Two dimensional models; UAV; Vegetation; Wireless communication systems; Wireless communications
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2907482

Modeling signal power path loss (SPPL) for deployment of wireless communication systems (WCSs) is one of the most time consuming and expensive processes that require data collections during link budget analysis. Radio frequency (RF) engineers mainly employ either deterministic or stochastic approaches for the estimation of SPPL. In the case of stochastic approach, empirical propagation models use predefined estimation parameters for different environments such as reference distance path loss PL<inline-formula> <tex-math notation="LaTeX">(d_{0}) </tex-math></inline-formula>(dB), path loss exponent <inline-formula> <tex-math notation="LaTeX">(n) </tex-math></inline-formula>, and log-normal shadowing <inline-formula> <tex-math notation="LaTeX">(X_{\sigma } </tex-math></inline-formula> with <inline-formula> <tex-math notation="LaTeX">N(\sigma ,\mu =0)) </tex-math></inline-formula>. Since empirical models broadly classify the environment under urban, suburban, and rural area, they do not take into account every micro-variation on the terrain. Therefore, empirical models deviate significantly from actual measurements. This paper proposes a smart deployment method of WCS to minimize the need for predefined estimation parameters by creating a 3-D deployment environment which takes into account the micro-variations in the environment. Tree canopies are highly complex structures which create micro-variations and related unidentified path loss due to scattering and absorption. Thus, our proposed model will mainly focus on the effect of tree canopies and can be applied to any environment. The proposed model uses a 2-D image color classification to extract features from a 3-D point cloud and a machine learning (ML) algorithm to predict SPPL. Empirical path loss models have received signal level (RSL) errors in the range of 6.29%-16.9% from the actual RSL measurements while the proposed model has an RSL error of 4.26%.