Using tensor model to extend least squares channel estimation problem for the intelligent reflecting surface assisted MIMO systems

Abstract

This paper proposes to extend two Least Squares (LS) channel estimation algorithms based on tensor signal model to MIMO systems supported by Intelligent Reflective Surfaces (IRS). These two algorithms exploit the tensor structure of the pilot signal to establish the cascaded channel estimation problem. The first algorithm extends the LS estimation based on exploiting the Khatri-Rao Factorization (KRF) structure of the cascaded MIMO channel, by solving subproblems approximating the 1-rank matrix. The second estimator is based on the Bilinear Alternating
Least Squares (BALS) algorithm, which is a simplified version of the Trilinear Alternating Least Squares (TALS) algorithm. In addition, this paper also presents the relationship between the MIMO channel parameters for the above estimation algorithms to be feasible. The simulation results show that the extended LS estimation methods based on the tensor signal model have improved performance compared with the conventional LS estimation.