A Fast Near-Field Beam Training Strategy based on Array Partition

In this paper, we study the efficient near-field beam training strategy for extremely large antenna array (ELAA). Unlike conventional far-field beam training, which only scans the angular domain, near-field beam training is more complex as it requires beam scanning over both angular and distance domains. To address the high time complexity of the twodimensional exhaustive beam search, we propose a fast nearfield beam training strategy that consists of the angle alignment (AA) stage and the distance alignment (DA) stage. We focus on the AA stage due to its more significant impact on training overhead. Drawing inspiration from several observations of the near-field beam patterns, an array partition-based multi-beam training scheme is proposed with much more reduced training overhead. Numerical results show that the proposed scheme can reduce over 95% training overhead compared to exhaustive search methods without sacrificing much of the achievable data rate, thus outperforming benchmark schemes in terms of the effective data rate.