Unified Block-Sparse Hybrid-Field Channel Estimation for XL-MIMO Systems

Exceptional hybrid-field channel estimation is crucial for providing accurate channel state information in extremely large-scale massive MIMO (XL-MIMO) systems, as channels are likely to involve both far- and near-field components. However, existing approaches, resorting to the use of channel sparsity, depend heavily on prior information about the proportion of far- and near-field paths in different sparse basis. To address this challenge, we propose an innovative hybrid-field channel estimation scheme based on an elaborating orthogonal least squares (OLS)-framework algorithm, which unifies the estimation into angular domain, thereby eliminating this potential dependence. Specially, a multiple block simultaneous OLS (MBSOLS) is first designed incorporating a multiple selection mechanism for reducing iterations, which can jointly utilize the simultaneously- and block-sparse channel characteristics. Meanwhile, the theoretical guarantees of MBSOLS in terms of the upper bound of reconstruction error are developed by employing mutual incoherence property (MIP). Based on these algorithmic outcomes, a fast MBSOLS (F-MBSOLS) algorithm is further elaborated via a sharp inverse matrix approximation formulation, substantiating the low-complexity feasibility of the proposed scheme. The hybrid-field channel is then efficiently estimated by F-MBSOLS in angular domain uniformly, regardless of additional prior information. Simulation results demonstrate that the elaborating scheme achieves outstanding accuracy under different hybrid-field structures, and remarkably low computational complexity to fulfill the demand of delay-sensitive vehicular communication.