Deep Learning-Based Uplink Timing Advance Estimation in Multiuser LEO Satellite Systems

Low Earth orbit (LEO) satellite communication (SatCom) is the key enabler to realize global coverage in future non-terrestrial networks (NTNs). However, the high mobility and wide serving area of LEO satellites cause fast timing-varying propagation delays and severe uplink timing misalignment, making the uplink timing synchronization particularly challenging in LEO SatCom. Existing works employ delay estimation and timing advance (TA) operations to align the arrival times of uplink signals. However, the fast time-varying nature of the delay is often overlooked, leading to inaccurate or outdated TA adjustments. Motivated by this, we propose a deep learning method incorporating a long short-term memory (LSTM) predictor to estimate uplink delays for TA in LEO SatCom uplinks. Simulation results show that the proposed method achieves low root-mean-square error (RMSE) and low bit error rate (BER), demonstrating its efficacy in addressing the timing misalignment in LEO SatCom uplinks.