Study on machine learning-based prediction of intensity fluctuations in horizontal optical propagation

This study investigates the potential of machine learning to predict signal power fluctuations in free-space optical communications, impacted by atmospheric fluctuations. Conducted over a 600-meter terrestrial link, optical reception intensities were recorded at a sampling rate of 10 kHz, then averaged to 1 kHz to reduce the computational load for subsequent machine learning processes. The study employed 80% of this data for training and 20% for validation. It was demonstrated that within the spatial coherence time of atmospheric turbulence, the fluctuations in received signal intensity can be accurately predicted using machine learning.