An Entropy-Based Adaptive DBSCAN Clustering Algorithm and Its Application in THz Wireless Channels

Terahertz (THz) communication has emerged as a highly promising technology in the field of sixth-generation (6G) communication systems. The understanding of propagation behavior, channel characteristics, and the development of a realistic channel model are essential prerequisites for THz communications. Notably, THz channels exhibit the distinctive feature of sparse clustering, which is a characteristic unique to THz channels. In this article, we propose an entropy-based adaptive density-based spatial clustering of applications with noise (EBA-DBSCAN) algorithm for efficient cluster analysis in THz communication channels. The EBA-DBSCAN algorithm combines the entropy method (EM) and the median absolute deviation (MAD) method to determine the clustering order and obtain an adaptive neighborhood radius for each multipath component (MPC). Extensive measurement campaigns are conducted in an indoor L-shaped hallway, covering the frequency range from 215 to 225 GHz. The clustering performance and time complexity of the proposed algorithm are comprehensively evaluated. Furthermore, we analyze the cluster parameters by considering the distribution characteristics of the surrounding environment. The clustering characteristics presented in this study significantly contribute to a better understanding of radio propagation and serve as a foundation for the development of efficient and accurate cluster-based channel models for 6G THz communication systems.