Efficient Incremental Variational Bayesian Learning-Based Near-Field Channel Estimation for XL-MIMO Systems

Extremely large-scale multiple-input multiple-Output (XL-MIMO) technique has recently emerged as a promising candidate to improve the capacity and spectral efficiency for future sixth-generation (6 G) systems. However, efficient channel estimation that considers the corresponding near-field effect of modeling the channel under spherical wavefront assumption is challenging. In this case, the high-storage and high-complexity requirements of sparse channel representation with angle-distance sampling are major barriers. For this reason, we develop a novel channel estimation scheme for XL-MIMO systems within variational Bayesian learning (VBL) framework, where the sparse channel representation in angular domain is adopted. Firstly, considering the property that channel sparsity in angular domain plays a decisive role for near-field channels, we formulate the sparse channel representation and recovery problem with just angular sampling, which differs from those schemes of adopting angle-distance sampling. However, this inaccurate angular sampling degrades the performance of existing channel estimation methods. To deal with it, we propose an incremental variational Bayesian learning (IVBL) channel estimation scheme that adaptively selects columns in the angular-domain sparse representation dictionary based on derived threshold conditions, effectively capturing near-field angular spread. This approach enables channel recovery through dynamic angulardomain dictionary matching and weight adjustment. Finally, simulation results are provided to show the superiority of the proposed channel estimation method in terms of pilot overhead and computational complexity.