Robust Channel Estimation for Massive MIMO Systems

Massive multiple-input multiple-output (MIMO) is one of the most promising tech-nologies in 5G systems. The base station (BS) is equipped with a much larger number of antennas in massive MIMO systems than those in the conventional single-input single-output (SISO) systems, making the dimension of the channels significantly larger. This phenomenon causes the implementation of the linear mini- mum mean square error (LMMSE) channel estimation, which generally involves the second-order statistics, e.g., the channel covariance matrix or the precision matrix (the inverse of the covariance matrix), a challenging issue to be solved. Meanwhile, the computational complexity of implementing the LMMSE channel estimation is also a significant issue. This thesis aims to propose robust channel estimation for the massive MIMO systems. Two major aspects are investigated in this thesis: the estimation of the channel covariance matrix and the precision matrix to improve the LMMSE channel estimation, as well as the complexity reduction for it in the massive MIMO systems. Firstly, a two-stage shrinkage covariance matrix estimator is proposed to improve the LMMSE channel estimator for massive MIMO systems. The LMMSE channel estimator is optimal in terms of the mean square error (MSE) when the channel statistics are known. However, such statistics are not perfectly known in real-world scenarios and need to be estimated. A two-stage algorithm is proposed to select the shrinkage factors to improve the robustness of the covariance matrix estimation when the number of samples varies, contributing to the overall channel estimation performance. Numerical results are provided to demonstrate the effectiveness of the proposed method to obtain covariance matrix estimation, improving the LMMSE channel estimator in massive MIMO systems.