Codebook Design and Hybrid Reference Signaling for AI/ML-Based Beam Management

초록

Beam management is essential due to the need for highly directional beamforming and the associated measurement overhead, which becomes increasingly significant in emerging frequency ranges such as the upper mid-band. To reduce this overhead, 3GPP has proposed AI/ML-based beam selection schemes that require careful design of a measurement subset known as Set B. This study investigates three strategies for configuring Set B and proposes a multi-rank reference signaling method for scenarios employing localized codebooks. Furthermore, we introduce a hybrid approach that integrates localized and uniform codebooks and demonstrate its effectiveness through performance evaluations. © 2025 IEEE.

키워드