An optimal scheme of two-timescale channel estimation framework for RIS-assisted wireless communications

The high pilot overhead associated with channel estimation (CE) is a major challenge in reconfigurable intelligent surfaces (RIS)-assisted wireless systems. A two-timescale CE framework has been proven to be able to significantly reduce the pilot overhead. This framework employs a coordinate descent-based iterative refinement (CDIR) optimization algorithm based on full-duplex base station (BS) assumptions, yet it suffers from high computational complexity. In this paper, we reconsider the optimization algorithm and two working modes along with full-duplex BS. Firstly, the CE equations are reformulated as a second-order nonlinear rank-one optimization problem after data matrix completion. Then, the received complex-valued pilot matrix is partitioned and represented in real form, and an optimization method based on principal eigenvector approximation is proposed. Lastly, a semi-closed-form expression is used to construct the relation between the received pilots and the channel parameters. Discussions and simulation results demonstrate that the proposed method exhibits lower computational complexity, lower pilot overhead, and acceptable estimation accuracy compared to various benchmark schemes. Especially, when the signal-to-noise ratio of the received pilots is less than 20 dB for the case of two pilot transmitting antenna, the estimation accuracy surpasses that of the referenced CDIR optimization algorithm.