Attention-Enhanced BiLSTM Network for Accurate Channel Estimation in High-Mobility OTFS Communications

With the rapid development of sixth-generation (6G) mobile communications, high-mobility scenarios are drawing attention, as Doppler shifts significantly impair reliable transmission. Orthogonal time frequency space (OTFS) modulation has emerged, owing to its robustness against Doppler effects. However, even with OTFS, rapidly varying channel state information continues to pose a major challenge for accurate channel estimation, as Doppler spread introduces severe time and frequency selectivity. Thus, we propose an attention-enhanced bidirectional long short-term memory (BiLSTM) network that leverages pilot features to perform accurate channel estimation, thereby ensuring reliable transmission in high-mobility environments. Extensive simulations validate the effectiveness of the proposed network.