Beam Training for Partially-Connected Structure in XL-MIMO Systems with Spatial Non-Stationarity

This paper presents a beam training for XL-MIMO systems operating in near-field non-stationary channels with a partially-connected architecture. The proposed method accounts for the spatial non-stationarity by selecting distinct beamforming codewords for each subarray, which are tailored to the different channels experienced by individual antenna parts. Simulation results demonstrate that the proposed beam training scheme outperforms conventional techniques in all SNR ranges, offering improved beamforming performance and better adaptation to the near-field and non-stationary environments of XL-MIMO channels.