CHARACTERIZATION OF URBAN ENVIRONMENTS FOR MILLIMETER WAVE AND TERAHERTZ PROPAGATION USING MACHINE LEARNING MODELS

Abstract

The communication technologies of millimeter waves and terahertz are transforming wireless systems as wireless networks become faster and more efficient. Enabling 6G applications, self-driving cars, real-time remote sensing, and ultra high-definition video streaming all become possible. However, mm-wave and THz signals still struggle with severe obstructions in urban areas. Buildings, cars, and even vegetation can block, reflect, scatter, or attenuate these signals. For effective communication, these environments must be understood, and models designed around them need to be created. Many traditional approaches to modeling lack sufficient flexibility to account for the changing conditions that are typical of urban areas. This research proposes a machine learning (ML) framework for analyzing and predicting signal propagation in various urban environments. We collected real-world data from various urban environments, performed feature extraction, and applied our framework by training and testing multiple machine learning (ML) algorithms. Prediction accuracy, scalability, and computing cost were evaluated, among other factors. Findings confirmed that ML models, unlike traditional models, can be trained to recognize the characteristics of an environment and predict them more reliably.

Additionally, this approach enables agile and flexible planning for next-generation wireless networks.This research aims to develop smart communication systems by integrating concepts from smart city environmental systems with signal propagation theory. In particular, designing smart city frameworks for real-time model expansion and validation across different geographic regions remains a work in progress.