Learning-Based Predictive Beamforming for UAV Communications With Jittering

Abstract

As a promising technique for realizing future wireless networks, unmanned aerial vehicle (UAV) communications have drawn numerous attentions. The performance of practical UAV communication systems is limited by the presence of inevitable jittering due to the inherent random wind gusts. The jittering introduces angle ambiguity which is challenging for aligning the information beams between the UAV-mounted base station (BS) and the user equipment (UE). This letter develops a learning-based predictive beamforming scheme to address the beam misalignment caused by UAV jittering. In particular, a deep learning approach is adopted to predict the angles between the UAV and the UE. By doing so, the UAV and the UE can prepare the transmit and receive beams in advance, which enables reliable UAV-based communication. Simulation results verify that the communication performance of the proposed scheme is robust to the presence of UAV jittering.