End-to-End Learning for Joint CSI Feedback and Prediction in FDD MIMO–OFDM Systems
IEEE Transactions on Wireless Communications · 2026 · 25 · 19034-19049
DOI: 10.1109/twc.2026.3702764 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Wireless Communications, 2026. Verfasst von Siyuan Tan und 3 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Gefördert von: Zhongxing Telecommunication Equipment Corporation (ZTE) Industry-University-Institute Cooperation Funds.
Abstract (Originalsprache)
Achieving high performance in Frequency Division Duplexing (FDD) massive Multiple-Input Multiple-Output (MIMO) systems is fundamentally challenged by channel state information (CSI) feedback overhead and channel aging, particularly in mobile environments. Conventional approaches address CSI feedback and prediction as separate, cascaded tasks, which is inherently suboptimal. This paper proposes a novel end-to-end learning framework that unifies the entire process from CSI estimation and feedback to latent-space prediction and downlink precoding. At the core of our framework is a new, computationally efficient channel prediction model, the Patched Multi-Layer SAMformer (PMSAM), which operates directly on the compressed latent representation of the channel. To enhance robustness and generalization, PMSAM is trained using Sharpness-Aware Minimization (SAM). Furthermore, our framework uniquely incorporates uplink Sounding Reference Signals (SRS) by exploiting weak channel reciprocity, using uplink information to refine downlink prediction accuracy. Extensive simulation results show that the proposed framework achieves lower prediction error and higher downlink spectral efficiency than representative cascaded feedback-and-prediction baselines under the considered feedback budget.
Förderung
- Zhongxing Telecommunication Equipment Corporation (ZTE) Industry-University-Institute Cooperation Funds
Zitieren
APA
Tan, S., Yuan, X., Duan, X., & Xu, X. (2026). End-to-End Learning for Joint CSI Feedback and Prediction in FDD MIMO–OFDM Systems. IEEE Transactions on Wireless Communications, 25, 19034-19049. https://doi.org/10.1109/twc.2026.3702764
Vancouver
Tan S, Yuan X, Duan X, Xu X. End-to-End Learning for Joint CSI Feedback and Prediction in FDD MIMO–OFDM Systems. IEEE Transactions on Wireless Communications. 2026;25:19034-19049. doi:10.1109/twc.2026.3702764
BibTeX
@article{Tan2026,
author = {Tan, Siyuan and Yuan, Xiaojun and Duan, Xiangyang and Xu, Xiaojing},
title = {End-to-End Learning for Joint CSI Feedback and Prediction in FDD MIMO–OFDM Systems},
journal = {IEEE Transactions on Wireless Communications},
year = {2026},
volume = {25},
pages = {19034--19049},
doi = {10.1109/twc.2026.3702764}
}
EUVIMED