Dynamic Hierarchical Beam Training via Gain Drop Detection for Near-Field Communications

We propose a dynamic hierarchical beam training algorithm to address the challenge of the large beam training overhead in near-field extremely large-scale array (XL-array) systems. Building on far-field hierarchical beam search, the proposed algorithm introduces a beam gain drop detection method to identify variations in beam gain across hierarchical layers caused by near-field effects. Following the beam gain drop detection, the algorithm estimates the user range by comparing the received power and dynamically adjusts the target distance. Through iterative beam gain drop detection and target distance refinement across layers, the algorithm achieves efficient beam training with significantly reduced overhead. Numerical results demonstrate that the proposed algorithm substantially lowers beam training overhead while maintaining reasonable and robust performance across various user scenarios, achieving practical feasibility for near-field communications.