Near-Field Beam Training Design for Stacked Intelligent Metasurface Assisted Communication Systems

The technique of stacked intelligent metasurfaces (SIMs) have emerged as a promising solution to enable advanced wave-domain processing with reduced hardware complexity. However, existing SIM research works are mainly based on the assumption of perfect channel state information (CSI), ignoring the critical challenge of channel acquisition. To address this problem, this paper proposes a novel near-field beam training framework for SIM-assisted systems. Specifically, we first establish an end-to-end near-field channel model of the SIM-assisted systems. Based on this model, we design a DFT based beam training scheme that performs a matched-filter search over distance, elevation, and azimuth to localize the user equipment (UE). The obtained position estimate is then utilized to construct the equivalent channel, which serves as the foundation of the subsequent beamforming algorithm by refining the digital combiner and the SIM phase shifts alternatively. Numerical results demonstrate that the proposed SIM-aided near-field beam training framework significantly improves spectral efficiency compared to the conventional schemes without SIMs.