Near-Field Channel Estimation via Dynamic Grids and Bayesian Inference

In this paper, we propose a dynamic grid sparse Bayesian inference (DGSBI) method for near-field channel estimation in extremely large-scale multiple-input multiple-output (XL-MIMO) systems. To address the off-grid problem of scatterers in the polar-domain channel, we represent the polardomain dictionary dynamically using angle and distance offsets and provide a sparse representation of the channel within this dynamic dictionary. By updating the grid through angle and distance offset variables, the proposed method significantly improves signal recovery performance in the sparse polar domain. The dictionary update process is modeled as a linearquadratic (LQ) optimization problem, and we derive an analytical solution for the update. Simulation results demonstrate that our method outperforms benchmark approaches in terms of estimation accuracy.