Millimeter Wave Path Loss Modeling for UAV Communications Using Deep Learning

Unmanned Aerial Vehicles (UAVs) and millimeter waves are pivotal technologies in the sixth-generation (6G) mobile communication systems. Effective path loss modeling for UAV-based millimeter wave communications is critical for rapid and accurate data transmission. Traditional methods, such as deterministic, empirical, and machine learning-based approaches, are commonly used. This paper presents a groundbreaking approach that harnesses the power of deep learning, specifically the Long Short-Term Memory (LSTM) algorithm, to predict path loss in UAV-based millimeter wave communications, with a particular focus on UAV-to-UAV scenarios. Our experimental results demonstrate the exceptional performance of our deep learning model, achieving a remarkable term root-mean-square error (RMSE) of only 1.98 dB when compared to measurement results in test scenarios. This remarkable outcome underscores the profound significance of employing deep learning methodologies in predicting path loss, surpassing the capabilities of traditional methods. By leveraging deep learning, we advance the field of UAV-based millimeter wave communication modeling, enabling more precise and efficient data transmission in 6G networks.