Scalable Transceiver Design for Multi-User Communication in FDD Massive MIMO Systems via Deep Learning

TL;DR

This work tackles scalable downlink design for FDD massive MIMO by jointly optimizing pilot transmission, channel feature extraction/feedback, and multi-user precoding with a single training phase. It introduces an RVQ-VAE with a hierarchical codebook for flexible feedback and an edge graph attention network (EGAT) for scalable precoding, complemented by a progressive training strategy. The framework demonstrates substantial sum-rate gains and robust generalization across varying feedback capacities and user counts on real-world channel data. The results indicate strong potential for deploying scalable, data-driven transceivers in practical FDD massive MIMO networks.

Abstract

This paper addresses the joint transceiver design, including pilot transmission, channel feature extraction and feedback, as well as precoding, for low-overhead downlink massive multiple-input multiple-output (MIMO) communication in frequency-division duplex (FDD) systems. Although deep learning (DL) has shown great potential in tackling this problem, existing methods often suffer from poor scalability in practical systems, as the solution obtained in the training phase merely works for a fixed feedback capacity and a fixed number of users in the deployment phase. To address this limitation, we propose a novel DL-based framework comprised of choreographed neural networks, which can utilize one training phase to generate all the transceiver solutions used in the deployment phase with varying sizes of feedback codebooks and numbers of users. The proposed framework includes a residual vector-quantized variational autoencoder (RVQ-VAE) for efficient channel feedback and an edge graph attention network (EGAT) for robust multiuser precoding. It can adapt to different feedback capacities by flexibly adjusting the RVQ codebook sizes using the hierarchical codebook structure, and scale with the number of users through a feedback module sharing scheme and the inherent scalability of EGAT. Moreover, a progressive training strategy is proposed to further enhance data transmission performance and generalization capability. Numerical results on a real-world dataset demonstrate the superior scalability and performance of our approach over existing methods.

Figures (8)

• Figure 1: Block diagram of the proposed DL-based framework, where the dependencies on $B$ for $\hat{\mathbf{v}}_k$ and $\mathbf{W}_K$ are omitted for the sake of notation simplicity.
• Figure 2: Edge representation of an multi-user MIMO system with $M=3$ and $K=2$. For edge $(1,2)$, the neighboring edges include orange dashed lines (connections from Antenna $2$ to other users) and green dashed lines (connections from User $1$ to other antennas).
• Figure 3: Structure of the proposed EGAT precoding design.
• Figure 4: An example of the ray-tracing simulation setup is illustrated. The gray objects represent buildings. The red cubes indicate the positions of single-antenna users, distributed across the simulation area. The green cube array marks the location of the BS.
• Figure 5: Sum-rate v.s. SNR with $M=128, L=32, K=6, B=30, N = 3$.

Theorems & Definitions (2)

• Remark 1
• Remark 2