An Improved Hierarchical Codebook Design Based on the Gradient Descent Method

The characteristics and demerits of the conventional hierarchical codebooks are investigated, which have been reported for an efficient beam search in millimeter wave massive multi-input multi-output systems. The followings are required for desirable hierarchical codebooks: every codeword should have constant modulus entries for feasible implementation, every codeword should force all the antenna elements to be active for preventing from total antenna power loss, and neighboring codeword beams should have no interference with each other for high SNR. However there is no conventional scheme satisfying all the requirements. In this paper, the gradient descent method, which is one of the iterative optimization approaches, is applied to derive the codebook meeting all the requirements. It is confirmed by computer simulations that the codebook obtained by the proposed method has larger achievable data rates than the conventional ones.

Paper

Full text

PDF

An Improved Hierarchical Codebook Design Based on the Gradient Descent Method

Semantic Scholar · Engineering · 2020

Abstract

The characteristics and demerits of the conventional hierarchical codebooks are investigated, which have been reported for an efficient beam search in millimeter wave massive multi-input multi-output systems. The followings are required for desirable hierarchical codebooks: every codeword should have constant modulus entries for feasible implementation, every codeword should force all the antenna elements to be active for preventing from total antenna power loss, and neighboring codeword beams should have no interference with each other for high SNR. However there is no conventional scheme satisfying all the requirements. In this paper, the gradient descent method, which is one of the iterative optimization approaches, is applied to derive the codebook meeting all the requirements. It is confirmed by computer simulations that the codebook obtained by the proposed method has larger achievable data rates than the conventional ones.

Similar papers

© 2026 NYSGPT2525 LLC