Robust Hybrid-Field Sparse Channel Estimation for Extremely Large-Scale MIMO Systems

Abstract

An extremely large-scale multiple-input multipleoutput (XL-MIMO) system is a promising enabler for sixthgeneration (6G) wireless communications. Its hybrid-field channel exhibits a transition from near-field spherical wavefronts to far-field planar wavefronts, with the boundary defined by the Rayleigh distance. This dual-wavefront characteristic complicates the construction of an accurate channel support set and undermines conventional sparsity assumptions due to the joint dependence on angular and polar parameters across discretized grids. To address these challenges, we propose a sparse-aware orthogonal matching pursuit (OMP)-based channel estimation method. The method integrates QR decomposition to ensure numerical stability of the atom dictionary and incorporates signal-to-noise ratio (SNR)-adaptive $\ell_{2}$-norm regularization to enhance estimation robustness against noise. Furthermore, we introduce a two-stage residual propagation strategy, wherein the far-field estimation residual serves as the initial residual for subsequent near-field refinement, thereby preserving estimation accuracy and mitigating performance degradation. A backtracking mechanism is employed to prioritize atoms with dominant correlation magnitudes, while dual halting criteria of residual energy thresholding and iteration count bounding are jointly designed to balance computational efficiency and reconstruction fidelity. Simulation results demonstrate that the proposed method achieves a significantly lower normalized mean squared error (NMSE) compared to the conventional OMP algorithm under identical conditions.