Dictionary Learning (DL)-based Sparse Cascaded Channel Estimation in IRS-assisted Terahertz MU-SIMO Systems With Hardware Impairments

This work conceives a sparse channel estimation (CE) scheme for multi-user (MU) intelligent reflecting surface (IRS)-aided Terahertz (THz) systems. The proposed framework also incorporates hardware impairments that arise due to manufacturing errors in practical THz systems, such as mutual coupling, irregular antenna spacing, and antenna gain/phase errors. A dictionary learning (DL) algorithm is proposed to learn the best sparsifying dictionary for an IRS-aided THz system in the presence of hardware impairments. The dictionary thus obtained is subsequently employed to leverage the sparsity inherent in the IRS-aided cascaded THz system toward channel estimation (CE). Simulation results corroborate our analytical findings and demonstrate the improved performance with respect to an agnostic scheme that ignores the non-idealities.