Statistical channel state information acquisition for massive MIMO communications

Abstract

This paper studies the statistical channel state information (CSI) acquisition in single-cell for massive multiuser multiple-input multiple-output (MIMO) system in the beam domain. Compared with the Bayesian channel estimation which models the channels by an independent and identically Gaussian-mixture (GM) distribution, we model the channels by an independent and non-identically Gaussian distribution exploiting the statistical channel characteristics, and provide the acquired statistical CSI for the approximate message passing (AMP) channel estimation that can reduce computational complexity. Meanwhile, we show that the statistical CSI for the user scheduling in the beam division multiple access (BDMA) transmission is the same as that of the AMP channel estimation. Due to the heavy orthogonal pilot overhead when serving a large number of user terminals (UTs) in massive MIMO system, the Bayesian channel estimation that don't need the orthogonal pilots is used to estimate the channel parameters in the statistical CSI acquisition. Simulation results show that the mean square error (MSE) of the AMP channel estimation using the estimated statistical CSI is better than that of the Bayesian channel estimation and the iteration number has a significant reduction. Besides, the user scheduling using the estimated statistical CSI can reach a good sum rate of the scheduled UTs.