Novel Codebook Design for Channel State Information Quantization in MIMO Rician Fading Channels With Limited Feedback

When a channel consists of a line-of-sight (LoS) path as well as non-LoS components, codebook design for channel state information (CSI) quantization is required to take account of both of them. However, the conventional codebook design requires infinitely many optimal codebooks corresponding to all possible Rician factors, which is impossible in practice. In this regard, we propose an effective codebook adaptive to any Rician factors, while guaranteeing comparable performance to the optimal codebook. Contrary to the conventional approaches, the adaptation to Rician factors suffices by sharing only a single common codebook between the transmitter and receiver. We first investigate the distribution of the angle between the channel vector and the LoS component, where the distribution depends on Rician factors that reflect the power ratio of LoS and non-LoS components. Driven by the analysis, we devise a band-structured non-homogeneous codebook and derive the upper bound of the quantization error of the proposed codebook. The design parameters of the proposed codebook are optimized to minimize the quantization error bound. Using an approximation, we also derive a tractable near-optimal solution of the parameters determining the proposed codebook. Numerical results exhibit that the proposed codebook substantially outperforms conventional methods and achieves near-optimal performance in terms of the average quantization distortion and average sum rate.