Path Loss Prediction Analysis for UAV to Maritime Channel Model using Meta Learning

Unmanned aerial vehicles (UAVs) have become increasingly essential in contemporary communication systems, necessitating precise channel models for UAV-to-maritime (U2M) interactions, which have remained challenging due to dynamic environmental conditions. This research has presented a hybrid machine learning framework combining meta-learning and long short-term memory (LSTM) networks to predict path loss (PL) for U2M non-line-of-sight (NLoS) scenarios. The proposed model incorporates critical environmental variables, including atmospheric attenuation, rain attenuation, and sea state dynamics, alongside operational parameters such as UAV altitude, antenna positioning, and propagation amplitude. By integrating meta-learning with LSTM, our framework captured temporal interdependencies and adapted to evolving channel conditions, enhancing robustness in dynamic maritime environments. The time-variant statistical properties, such as cumulative distribution function (CDF) and PL fluctuations, are analyzed and analytically derived to refine fading statistics, delay spread, and predictive accuracy. The results of the proposed framework are compared with theoretical results and existing models, such as the International Telecommunication Union (ITU) and the 3rd Generation Partnership Project (3GPP). Finally, the outcomes of the proposed model show that the simulation results align substantially with the corresponding theoretical results.