Graph-Regularization-Based Low-Rank Matrix Completion for Multi-User mmWave 3D-MIMO Channel Estimation

Three-Dimensional Multiple-Input Multiple-Output (3D-MIMO) system can effectively improve the spectrum efficiency(SE) and user experienced data rate for future wireless communication systems. However, the dramatic increase of users and antennas brings problems such as decreased estimation accuracy and increased complexity to 3D-MIMO channel estimation. This paper proposes a graph-regularization-based channel estimation approach, which has higher accuracy and lower complexity. Besides taking advantage of the inherent low-rankness in the mmWave MIMO channel, the approach also exploits the correlation between multiple users as side information for the estimation procedure. The resulting norm minimization problems is faster optimized via the Semi-Proximal Symmetric Alternating Direction Method of Multipliers(SS-ADMM) algorithm. Simulation results demonstrate that our proposed scheme can recover channels effectively and attain a significant performance improvement over comparison methods.