Tensor-Based Near-Field Channel Estimation for XL-RIS-Assisted Terahertz Systems

In terahertz (THz) communication systems, extremely large scale arrays can effectively compensate for the limited communication distance problem. In this article, we consider the near-field channel estimation (CE) problem for an extremely large reconfigurable intelligence surface (XL-RIS)-assisted multi-user THz communication system. We first construct a near-field channel model based on a second-order Fresnel approximation derivation. Utilizing the spatial structure of the derived channel model, we sample the covariance matrix of the received signals. Then, we propose a tensor decomposition-based algorithm to estimate the angular parameters, and establish a truncated singular value decomposition (T-SVD) algorithm for the distance estimation. In the end, we estimate the path losses through the least squares (LS) method and recover the complete channel. Moreover, to further reduce the computational overhead, we construct a low-complexity tensor completion-based scheme for the angular parameters’ estimation. Simulation results indicate that the proposed tensor-based CE schemes outperform the conventional subspace-based approaches in terms of accuracy and computational complexity.