GroupMixFormer: Another Blessing of DL-Based Massive MIMO Hybrid Beamforming

Ye Zeng, Li Qiao, Hengwei Zhang, Ziqi Han, Kuiyu Wang, Yang Huang, Zhen Gao*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The hybrid beamforming plays a critical role in massive multiple-input multiple-output (MIMO), addressing the high hardware costs associated with fully-digital beamforming. The design of hybrid beamforming, particularly satisfying the constant modulus constraint for analog beamforming, presents a complex non-convex problem. In this paper, we propose a GroupMixFormer-based hybrid beamforming network (GMF-HBN) for multi-user scenarios using orthogonal frequency di-vision multiplexing (OFDM) with both perfect and imperfect channel state information (CSI). Simulation results show a significant performance improvement of our proposed GMF-HBN with perfect CSI compared to the Transformer-based and also demonstrate robustness with imperfect CSI.

Original languageEnglish
Title of host publicationInternational Conference on Communications in China, ICCC Workshops 2024
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages470-474
Number of pages5
ISBN (Electronic)9798350377675
DOIs
Publication statusPublished - 2024
Event2024 IEEE/CIC International Conference on Communications in China, ICCC Workshops 2024 - Hangzhou, China
Duration: 7 Aug 20249 Aug 2024

Publication series

NameInternational Conference on Communications in China, ICCC Workshops 2024

Conference

Conference2024 IEEE/CIC International Conference on Communications in China, ICCC Workshops 2024
Country/TerritoryChina
CityHangzhou
Period7/08/249/08/24

Keywords

  • deep learning (DL)
  • GroupMixFormer
  • hybrid beamforming
  • multiple-input multiple-output (MIMO)

Cite this