Codebook-Based Trellis-Coded Quantization Scheme Using K-Means Clustering for Massive MIMO Systems

Abstract

This paper introduces a codebook-based trellis-coded quantization (TCQ) approach utilizing K-means clustering, designed specifically for massive multiple-input multiple-output systems. The proposed TCQ scheme follows the structure of conventional TCQ, quantizing channels through a trellis and source constellation to minimize latency. The performance of channel quantization significantly depends on the chosen source constellation. The conventional TCQ scheme, using a fixed set of constellations as source constellations, imposes limitations on beamforming gain performance. In our proposed TCQ scheme, we employ a codebook to utilize a set of constellations tailored to the channel environments as the source constellation, aiming for enhanced beamforming gain performance. This approach leads to a slight increase in feedback overhead. The codebook’s constellation sets are generated by determining centroids through K-means clustering of channel elements and mapping them to high-order quadrature amplitude modulation constellations. Simulation results demonstrate that our proposed TCQ scheme, with a comparable computational complexity, shows the improved average beamforming gains compared to the conventional TCQ scheme.

Keywords

• Limited feedback
• massive MIMO system
• trellis-coded quantization (TCQ)
• URLLC
• K-means clustering