IEEE Access (Jan 2024)

RF Chain-Wise Clustering Schemes for Millimeter Wave Cell-Free Massive MIMO With Centralized Hybrid Beamforming

  • Shunsuke Kamiwatari,
  • Issei Kanno,
  • Takahiro Hayashi,
  • Yoshiaki Amano

DOI
https://doi.org/10.1109/ACCESS.2024.3360716
Journal volume & issue
Vol. 12
pp. 19682 – 19693

Abstract

Read online

In this paper, we propose clustering schemes appropriate for a cell-free massive multi-input multi-output (CF mMIMO) system with centralized hybrid beamforming (BF) in the millimeter wave (mmWave) band, with the aim of reducing computational complexity and achieving scalability at the central processing unit (CPU) while ensuring system performance. Conventional access point (AP)-wise clustering schemes have been proposed for a decentralized architecture that mitigates inter-user interference at each AP through local digital BF and analog BF of each radio frequency (RF) chain to efficiently improve performance. However, in a centralized architecture, in which the digital BF weights is designed at the CPU while considering all RF chains of all APs together. In cases where each AP employs multiple RF chains forming different analog beams, the coupling loss (the sum of the path loss and BF gain of the analog beam) of each RF chain may differ even between the same AP and UE due to differences in the analog BF gain. AP-wise clustering is not well suited because RF chains with a high coupling loss (the sum of the path loss and BF gain of the analog beam) may be included among the APs of the cluster assigned to each UE. In the method proposed in this paper, clusters are formed on a per-RF-chain basis (RF chain-wise clustering). In this way, RF chains with lower coupling loss can be selected for each UE regardless of to which AP an RF chain belongs when forming a cluster for each UE. Moreover, the existing clustering schemes do not account for the amount of inter-cluster interference. Accordingly, we propose a cluster recombination scheme to effectively mitigate inter-cluster interference. Then, by combining RF chain-wise clustering and cluster recombination, a higher received signal power can be achieved while mitigating inter-cluster interference. Through simulation-based evaluations, we show that the proposed RF chain-wise clustering and cluster recombination schemes can achieve superior spectral efficiency while effectively reducing the complexity of centralized digital BF.

Keywords