Sparse Channel Estimation for Intelligent Reflecting Surface Assisted Massive MIMO Systems

Intelligent reflecting surface (IRS) has emerged as a promising technology for improving the spectrum and energy efficiency of next-generation wireless communications. Accurately acquiring channel state information (CSI) of the IRS-assisted wireless system is an essential task for reaping the passive beamforming gain of IRS. However, it is quite difficult to directly obtain the CSI of the IRS-assisted wireless system due to the inability of signal processing at the IRS. In this paper, we investigate the downlink channel estimation problem for IRS-assisted massive MIMO systems. We first present a new sparse recovery problem formulation for the cascaded downlink channel estimation, which exploits the sparsity of massive MIMO channels at the base station (BS) side and adopts the sub-surface idea to significantly reduce computational complexity and training overhead. In this case, the corresponding sparse signal recovery problem exhibits a row-sparse feature but is additionally affected by a coupling matrix. It is challenging to recover the row-sparse matrix with the traditional sparse representation methods, as the elements in the row-sparse matrix are highly coupled with each other. To meet the challenge, we employ a hybrid approximate message passing (AMP) framework for the cascaded downlink channel estimation, which consists of a part of expectation propagation approximation (EPA) and a part of two-level generalized approximate message passing (GAMP). We illustrate that combining EPA and GAMP into a hybrid framework can make up for their respective shortcomings. Numerical simulation results verify the superiority of the proposed method.

Automated summary

This paper investigates the downlink channel estimation problem for IRS-assisted massive MIMO systems. A new sparse recovery problem formulation is proposed, utilizing the sub-surface idea to reduce computational complexity. A hybrid approximate message passing framework is employed to recover the row-sparse matrix, and the superiority of the proposed method is verified through numerical simulations.