Geographical Clustering of Path Loss Modeling for Wireless Emulation in Various Environments

• Published in: 2022 16th European Conference on Antennas and Propagation (EuCAP) pp. 1 - 5
• Main Authors: Nagao, Tatsuya, Hayashi, Takahiro
• Format: Conference Proceeding
• Language: English
• Published: European Association for Antennas and Propagation 27.03.2022
• Subjects: clustering; Clustering methods; Computational modeling; Data models; deep learning; Emulation; Machine learning; path loss modeling; Radio propagation; Wireless communication
• DOI: 10.23919/EuCAP53622.2022.9769198

A testbed that emulates wireless communication in a virtual space is needed to efficiently validate new wireless communication systems in large-scale and various environments. For realistic wireless emulation, it is necessary to simulate the radio propagation characteristics site-specifically and accurately. Modeling methods that are based on machine learning using map data have been proposed. However, the use of a single modeling approach in various propagation environments may result in lower accuracy in some areas. Additionally, in single modeling in an extensive area, issues regarding computational cost and model-tuning efficiency are encountered. This paper proposes a method for clustering map data and for modeling on each cluster using machine learning. This method optimizes the modeling area according to the geographical features and improves the accuracy and computational efficiency. Evaluation results that are obtained using measurement data showed that the proposed method improves the accuracy and increases the speed by approximately 30%.