MUSIC-Based Channel Estimation with Adaptive Reconfiguration of Diagonal RIS

We consider a cellular system including a passive reconfigurable intelligent surface (RIS), without sensing or data processing capabilities. We focus on the problem of channel estimation for transmissions in the millimeter-wave band, where a small number of paths describes the propagation, each charac-terized by angles of arrival (AoAs), angles of departure (AoDs), and power gains. Starting from a recent algorithm here denoted MUSIC-based channel estimation (MBCE) we overcome some of its performance and technological limitations. In particular, we introduce the practical constraint that RIS elements can only be controlled in phase and cannot be switched off. This prevents an estimation procedure of the MBCE and we resort to a new linear minimum mean square error (LMMSE) approach for the estimation of path angles at the RIS. We further enhance channel estimation by making it iterative, and at each iteration we a) refine the estimation of AoA and AoD at the base station (BS) and user equipment (UE) and b) optimize the RIS configuration to be used in the next iterations. Numerical results confirm that the proposed channel estimation with adaptive RIS configuration (CEARC) approach improves the channel estimate over MBCE.