Learning-Aided Blind Beam Adaptation for UAV Communication Systems With Jittering

We propose an intelligent blind beam tracking and adaptation scheme for air-to-ground (A2G) communication systems. Our approach autonomously adjusts to dynamic channel conditions induced by UAV mobility without relying on explicit channel information or pilot signals during beam tracking. Addressing potential inaccuracies from channel impairments, including the jittering effect, we introduce two robust beam adaptation strategies leveraging the estimated probability mass function (PMF) from a deep learning model. The probability-based beam adaptation with weighted sum (PBA-WS) balances beam gain between primary and secondary peaks to prevent severe drops in case of estimation inaccuracies. The probability-based beam adaptation with received signal samples (PBA-RS) selects the beamformer with the highest gain to the previous received sample. Simulation results demonstrate the effectiveness and robustness of our proposed scheme, particularly in high channel variability environments.