Iterative Reweight Super-Resolution Channel Estimation in Hardware-Impaired Hybrid Precoded MM-Wave Massive MIMO systems using Dual SVD and Marquardt’s Global Search

Abstract

In this paper, a three-dimensional geometric channel model characterized by channel gain, angle of arrival, and departure (AoA/AoD) is used to develop a non-ideal and more realistic system representation that accounts for additional signal perturbations other than additive white Gaussian noise (AWGN). Hinged on this, compressive sensing (CS) based channel estimation technique is proposed for millimeter- wave (MM-Wave) massive multiple-input multiple-output (MIMO) systems with inherent challenges due to hardware impairment (HI). The proposed estimator named dual singular value decomposition (SVD) and Marquardt, abbreviated as DSM estimator aims at tolerating complexity and improving channel estimation accuracy at reasonable trade-offs in the system. The normalized mean square error (NMSE) performance of the proposed channel estimation scheme, with modified Marquardt’s global minimum search, achieves higher accuracy at below 0 dB (LoS)/-3 dB (NLoS) signal-to-noise ratio (SNR) for uniform linear array (ULA) configuration and performs better for uniform planar array (UPA) configuration. A low- rank range of unit difference is adopted for matrix decomposition twice (dual SVD) per SNR in ULA/UPA setting. The proposed estimator’s result shows early convergence during the simulation and a linearly scaled complexity in comparison with one of the earlier proposed schemes in literature in particular the super-resolution channel estimation with gradient descent optimization.