Near-Field Channel Estimation for RIS-Aided Industrial IoT Systems

Internet of Things (IoT) has emerged as a key application domain in modern wireless communication systems, where its effectiveness heavily relies on the accuracy of channel state information (CSI). The deployment of large-scale reconfigurable intelligent surfaces (RIS) makes near-field effects non-negligible, bringing great challenges to channel estimation in RIS-assisted IoT scenarios. Existing near-field estimation methods often suffer from modeling errors due to Fresnel approximation or excessive computational complexity arising from dense polar-domain sparse representation. In this paper, we propose a novel RIS-aided near-field channel estimation framework tailored for IoT environments. We first present an accurate sparse channel model that captures the true spherical wavefront propagation and thus eliminates the modeling inaccuracies introduced by Fresnel approximation. Leveraging this model, we further incorporate IoT environmental context to derive a novel dimensionality-reduced sparse representation. Subsequently, we devise a fast sparse Bayesian learning (SBL)-based sparsity recovery scheme with coarse off-grid refinement, and embed generalized approximate message passing (GAMP) to significantly reduce computational complexity. Simulation results demonstrate that the proposed method achieves high estimation accuracy with reduced complexity, owing to the dimensionality-reduced sparse representation and the fast GAMP-embedded SBL framework. These advantages make it highly suitable for RIS-assisted IoT systems.