Multi-Head DNN-Based Federated Learning for RSRP Prediction in 6G Wireless Communication

In the realm of wireless communications, accurate Radio Signal Received Power (RSRP) prediction serves as the foundation for improving user experience and optimizing network efficiency and reliability. With the deep integration of Artificial Intelligence (AI) technology and the wireless communicatio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE access 2024, Vol.12, p.97533-97543
Hauptverfasser: Yu, Menghan, Xiong, Xiong, Li, Zhen, Xia, Xu
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In the realm of wireless communications, accurate Radio Signal Received Power (RSRP) prediction serves as the foundation for improving user experience and optimizing network efficiency and reliability. With the deep integration of Artificial Intelligence (AI) technology and the wireless communication network, Federated Learning (FL) is considered as a promising approach for enhancing RSRP prediction while protecting user data privacy in the upcoming of 6G network. However, in practice, the heterogeneity of User Equipment (UE) environments and the limitations of UE communication bandwidth and computational capabilities can lead to poor model performance and inefficient model interactions in FL. To address these challenges, this paper proposes a Multi-head DNN based FL algorithm for RSRP prediction. The experimental results show that the proposed algorithm can enhance both RSRP prediction performance and communication efficiency.
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2024.3427694