Open AccessGreedy hybrid beamforming for multiuser MmWave MIMO systems
Author(s) -
Li Xiaohui,
Lin Yingchao,
Yang Xu,
Pu Wenjuan,
Meng Meimei,
Hei Yongqiang
Publication year - 2017
Publication title -
iet communications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.355
H-Index - 62
eISSN - 1751-8636
pISSN - 1751-8628
DOI - 10.1049/iet-com.2017.1000
Subject(s) - computer science , computational complexity theory , beamforming , greedy algorithm , singular value decomposition , baseband , matrix decomposition , matrix (chemical analysis) , algorithm , mimo , qr decomposition , dimension (graph theory) , mathematical optimization , mathematics , telecommunications , eigenvalues and eigenvectors , bandwidth (computing) , physics , materials science , quantum mechanics , pure mathematics , composite material
Due to the high cost and power consumption, it is impractical to perform full‐digital beamforming at baseband in millimetre‐wave systems. Thus, as a cost‐effective alternative, a hybrid beamforming (HBF) is introduced in this study. Since the HBF can be designed as a matrix factorisation problem, an effective greedy algorithm with low complexity for multiuser scenes is proposed to approach the full‐digital matrix. Firstly, initialisation for partial columns of the analogue matrix is enabled by extracting the phases of the full‐digital scheme. Then the singular value decomposition in low‐dimension matrix can be used to obtain the rest columns of the analogue matrix. Finally, based on effective channel, the digital matrix can be easily designed. Therefore, the authors' proposed algorithm has low complexity and only needs fewer iterations. Furthermore, by handling the inter‐user interference, the full‐digital matrix needed for the initialisation in analogue domain can be approximated to an orthogonal matrix, which can further reduce computational complexity. Simulation results demonstrate that the proposed greedy algorithm performs very close to the optimal unconstrained solution and enjoys low complexity.