Robust Free Space Optical Communication: Leveraging Hermite-Gaussian Modes and Derivative-Based Features

Free space communication plays a crucial role in the integration of terrestrial and non-terrestrial networks by providing high-speed, reliable backhaul links. It enables connectivity between ground stations and aerial or space-based nodes, such as satellites, high-altitude platforms (HAPs), and unmanned aerial vehicles (UAVs). In this paper, we present an AI-enhanced classification framework that analyzes Hermite-Gaussian (HG) modes for seamless network integration. Our system achieves high-accuracy identification of four distinct HG modes (HG00, HG01, HG02, and HG03) using experimentally measured Q-factor (Q) and Signal-to-Noise Ratio (SNR) parameters. The proposed methodology introduces a novel multi-stage approach that combines advanced feature engineering through first- and second-order discrete derivatives of Q and SNR measurements, intelligent dimensionality reduction via Principal Component Analysis (PCA), and adaptive k-Nearest Neighbors (KNN) classification. By implementing Kernel PCA alongside our derivative-based feature extraction technique, we demonstrate a significant improvement in classification accuracy from 94% to 98%, while achieving a 50% reduction in prediction latency. Experimental results show that our framework enables robust real-time mode identification in hybrid network architectures, making it particularly effective for applications requiring interaction between terrestrial and non-terrestrial systems.