Model-Based Structured Covariance-Aided Channel Estimation for Massive MIMO Systems

We consider the problem of downlink channel estimation in massive multiple-input multiple-output (MIMO) systems, where both transmitter (Tx) and receiver (Rx) are equipped with uniform planar array (UPA). The acquisition of channel state information (CSI) is important for subsequent signal detection and beamforming. To obtain the CSI accurately, the behavior of minimum mean-squared error (MMSE) channel estimator is studied, the main difficulty of which is the recovery of channel covariance matrix. In this paper, the channel covariance matrix is formulated under kronecker channel model, and the correlation of different antennas is formulated by the classical exponential model. Based on this model, the channel covariance matrix can be represented as a parameterized matrix, and the recovery of channel covariance matrix can be converted to an optimization problem of two parameters. An explicit solution is provided as channel estimator for independent identically distributed (i.i.d.) channels, while an alternate optimization algorithm is proposed to iteratively estimate the channel covariance matrix and the channel as well for correlated channels. The proposed algorithms improve the performance in exchange of computational complexity, and do not require extra pilot overhead. Simulation results are provided to illustrate the effectiveness of our proposed methods.