Cluster-Based Low-Complexity Codebook Design for Hierarchical Beam Training in XL-MIMO

This paper proposes a cluster-based low-complexity codebook design scheme for hierarchical near-field beam training in extremely large-scale MIMO(XL-MIMO), referred to as the cluster hierarchical beam training (CHB). Specifically, to reduce the codebook dimensionality while preserving essential angle and distance information, the proposed CHB scheme employs a cluster-based approach to identify cluster centers as new polar-domain sampling points. By doing so, CHB significantly reduces the complexity of codeword generation. The generated codewords are then applied to hierarchical beam training, thereby further decreasing the associated training overhead. Finally, simulation results confirm that CHB reduces complexity while offering comparable or even superior performance to other codebook-based near-field beam training schemes.