Two-Timescale Channel Estimation Scheme for Double-IRS-Assisted Massive MIMO System

‍ ‍In a massive multiple-input multiple-output （MIMO） communication system assisted by double intelligent reflecting surfaces （IRSs）， the channel estimation to obtain channel state information （CSI） requires significant pilot overhead. Considering that the base station （BS）-IRS and IRS-IRS channels have high-dimensional and quasi-static characteristics， while the IRS-user equipment （UE） channel has low-dimensional and time-varying characteristics， this study investigated a channel-estimation scheme based on two timescales. First， for the quasi-static channel， the base station sends pilots and estimates the BS-IRS and IRS-IRS channels using the coordinate descent algorithm and least square method. Then， based on the quasi-static channel-estimation results， the least square method is used to estimate the IRS-UE channel. Simulation results showed that the scheme proposed in this paper could obtain an estimated normalized mean square error that was nearly 10 dB lower than that of the comparison scheme， with only half of the pilot overhead of the comparison scheme. Thus， a higher channel-estimation accuracy was achieved with less pilot overhead.