Transformer-Empowered Predictive Beamforming for Rate-Splitting Multiple Access in Non-Terrestrial Networks

Existing Rate-Splitting Multiple Access (RSMA) techniques offer a promise for Non-Terrestrial Networks (NTNs) by managing interference and ensuring reliable data transmission. However, precoder design remains a crucial bottleneck, demanding accurate Channel State Information (CSI) feedback and compl...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on wireless communications Jg. 23; H. 12; S. 19776 - 19788
Hauptverfasser: Zhang, Shengyu, Zhang, Shiyao, Yuan, Weijie, Quek, Tony Q. S.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.12.2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Accuracy
Array signal processing
Artificial neural networks
Beamforming
Complexity
Convolutional neural networks
Data transmission
Design optimization
Feedback
Harnesses
Interference
Low earth orbit satellites
Machine learning
Multiple access
Non-terrestrial networks (NTNs)
predictive beamforming
Predictive models
Protocols
rate-splitting multiple access (RSMA)
Resource management
Satellites
Splitting
Transformers
ISSN:1536-1276, 1558-2248
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Existing Rate-Splitting Multiple Access (RSMA) techniques offer a promise for Non-Terrestrial Networks (NTNs) by managing interference and ensuring reliable data transmission. However, precoder design remains a crucial bottleneck, demanding accurate Channel State Information (CSI) feedback and complex optimization, which are challenging in practical deployment. Motivated by this, this paper proposes a novel Deep Learning (DL)-based method to predict the precoder design from the historical CSI directly. In particular, we first establish a predictive beamforming protocol for precoder design using historical CSI, bypassing the need for constant feedback and reducing complexity. Subsequently, we formulate a general problem for precoder design, with the Weighted Ergodic Sum Rate (WESR) serving as the objective function. Solving this problem is particularly challenging due to the dynamic nature of wireless channels in NTNs. To address this, we designed a fusion model, named TranCN, which harnesses the strengths of Transformers and Convolutional Neural Networks (CNNs) to extract spatial-temporal features from historical CSI, thereby enhancing precoder performance. Simulation results demonstrate that our predictive beamforming scheme enables RSMA to adapt to dynamic channel conditions using historical CSI, surpassing baseline methods and improving data transmission resilience.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1536-1276
1558-2248
DOI:10.1109/TWC.2024.3486673