Parallel Forecasting of LEO Satellite Downlink Channels Using a CNN-Transformer Architecture

Low Earth Orbit (LEO) satellite constellations have been widely recognized as a key component of sixth-generation (6G) wireless communications due to their great potential in enabling integrated space-air-ground networks. However, the high mobility of LEO satellites introduces significant challenges to the real-time and accurate acquisition of channel state information (CSI), which is critical for adaptive technologies in satellite communication systems. In this paper, we propose a deep learning-based channel prediction approach that combines Convolutional Neural Networks (CNN) with the Transformer architecture to achieve parallel prediction of future CSI from historical data. Simulation results show that the proposed method outperforms existing techniques, demonstrating enhanced accuracy and robustness, thus providing an effective solution for channel prediction in satellite communication systems.