Two-Timescale Channel Estimation for RIS-Assisted Near-Field Communication

We present a novel, two-timescale scheme to estimate the channel between a user and a multi-antenna base station (BS) in a reconfigurable intelligent surface (RIS)-assisted system that operates in the near-field regime. We obtain the maximum likelihood estimates of the large-scale parameters of the user-to-RIS channel, namely distance and directional cosine, only once in every block of contiguous coherence intervals. The small-scale, complex fading gain of the user-to-RIS channel in a coherence interval is then estimated using the above estimates and the pilot signals transmitted in that interval. These estimates, along with the deterministic RIS-to-BS channel gain, are combined to determine the cascaded channel gain vector from the user to the BS. We also present a minorization-maximization-based algorithm to determine the RIS configuration that maximizes the beamfocusing gain for data transmission. Our scheme results in a lower mean square error and a higher beamfocusing gain compared to several benchmarking schemes.