Transmit Antenna Selection Using Gated Recurrent Units for UAV-to-LEO Satellite Communications

In this study, we propose a transmit antenna selection (TAS) method based on gated recurrent units (GRUs) for communication between unmanned aerial vehicles (UAVs) and low Earth orbit (LEO) satellites. The GRU network can accurately predict the post-processed signal-to-noise ratio (pSNR) for each antenna by capturing the multiscale, time-varying channel quality of the downlink (i.e., from the LEO satellite to the UAV). Based on predicted pSNR, the UAV can select the appropriate antenna indices for subsequent uplink transmissions (i.e., from the UAV to the LEO satellite). Consequently, the pSNR can be improved, thereby improving communication performance between the UAV and the LEO satellite. The simulation results demonstrate that the proposed GRU-based TAS method outperforms existing TAS methods, including a convolutional neural network and long short-term memory-based TAS methods, with significantly reduced computational complexity.