Two-Timescale, Low-Overhead Parametric Channel Estimation for RIS-Assisted Near-Field Communication and Its Rate Implications

Abstract

New techniques are required to estimate the user-to-reconfigurable intelligent surface (RIS) near-field channel due to the distance-dependency of its steering vector. The estimate is needed to determine the RIS configuration and the combining vector of the multi-antenna base station (BS) receiver. We present a novel, low-overhead, two-timescale scheme to estimate gains of a multi-cluster user-to-RIS channel and a cascaded user-RIS-BS channel. It has low control and signaling overheads, and achieves a lower mean square error and a higher achievable rate than conventional one-timescale and far-field schemes. The large-scale parameters, namely distances and directional cosines, are estimated only once in a block of contiguous coherence intervals, while the small-scale, complex fading gains are estimated using the above estimates and the pilot signals transmitted in the same interval. We derive the Cramér-Rao lower bound for estimating the large-scale parameters. For a single-cluster channel, we also derive an insightful, novel expression for the achievable rate that accounts for noisy channel estimates, imperfect RIS configuration and combining vector, and training overhead. We derive the optimal RIS configuration and pilot and data powers that maximize the rate. Our numerical results present new insights about the ratio of the optimal pilot and data powers and the optimal number of pilots as a function of the coherence interval duration and the block size.