Att-GRU: An Attention-Enhanced Gated Recurrent Unit for Channel Prediction in Deep Space Communications

Lianning Cai, Guanjun Xu, Dusit Niyato

Abstract

Precise channel state information (CSI) is crucial to ensure the reliability of deep-space communications. However, severe Doppler shifts and solar scintillation during superior solar conjunction lead to rapidly time-varying channels, making the obtained CSI outdated. To solve this problem, we propose an attention-enhanced gated recurrent unit (Att-GRU)-based channel prediction scheme for deep-space communications. We first establish a time-varying deep-space channel model affected by Doppler shift and solar scintillation. Subsequently, we integrate the attention mechanism into the GRU and propose an Att-GRU-based channel prediction scheme to capture the temporal dependencies in the deep-space channel. By dynamically assigning weights to different time steps, the proposed scheme further improves the model's ability to predict complex channel variations. Finally, simulation results show that the proposed Att-GRU achieves up to 16.5 % normalized mean squared error (NMSE) reduction compared with the GRU model.
BibTeX
@article{cai2026att,
  author = {Lianning Cai and Guanjun Xu and Dusit Niyato},
  title = {{Att-GRU: An Attention-Enhanced Gated Recurrent Unit for Channel Prediction in Deep Space Communications}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2026},
  doi = {10.1109/tvt.2025.3637170},
}

Loading related content...

Figure