One GAMP-Based Learning Scheme for the Time-Varying Massive MIMO Channels

This paper proposes a novel scheme for learning the channel statistics of the time- varying massive MIMO network. In particular, the effects of the quantization at the receiver are considered. Firstly, we formulate the massive MIMO channel as a simultaneously time-varying sparse signal model through virtual channel representation (VCR) and first order auto regressive (AR) model. Then, we propose a sparse Bayesian learning (SBL) framework to learn the model parameters of the sparse virtual channel. To avoid the unacceptable complexity, we apply the expectation maximization (EM) algorithm to achieve the approximate solution. Specifically, the factor graph and the general approximate message propagation (GAMP)-based message passing algorithms are used to compute our wanted posterior statistics in the expectation step. After that, the non-zero supporting vector of virtual channel is obtained from channel statistics by a k-means clustering algorithm. Finally, we demonstrate the efficacy of the proposed schemes through simulations.