Deep Unfolding of Iterative Precoding-Based Null-Space Expansion for Multi-User Massive MIMO in Time-Varying Channel

This paper proposes a novel low-complexity precoding weight for downlink of multi-user massive multiple-input multiple-output (MU-mMIMO) system in time-varying channel environments. Inter-user interference (IUI) suppression performance in the MU-MIMO system deteriorates due to the mismatch between the channel state used for precoding weight derivation and that during signal transmission. Leveraging the massive antenna degrees of freedom (DOFs) in mMIMO, null-space expansion (NSE) was previously conceived to improve IUI suppression performance by steering multiple nulls toward undesired user equipments (UEs), even in high-mobility environment. However, NSE has severe challenges, such as high computation complexity caused by the increasing nullification and near singular matrix structure of the Gram matrix due to high time correlation. Considering these challenges, this paper proposes Jacobi matrix-based iterative low-complexity NSE weight design. Focusing on the Gram matrix’s block diagonal structure in NSE, this paper presents a Jakes’ model-based approximation and Ridge parameter optimization using deep unfolding (DU) to enhance the convergence of the inverse matrix. The effectiveness of the proposed scheme is demonstrated through computer simulations.