Precoder Design in Multi-User FDD Systems with VQ-VAE and GNN

2025-10-10

AI Summary

In FDD multi-user systems, conventional precoder design lacks robustness to channel statistics modeling, and traditional Gaussian Mixture Models (GMMs) suffer from exponential growth in component count with increasing feedback bits. Method: This paper proposes an end-to-end joint learning framework integrating Vector Quantized-Variational Autoencoders (VQ-VAEs) and Graph Neural Networks (GNNs), co-optimized with learnable pilots. VQ-VAEs replace GMMs to yield compact, scalable channel statistical representations; GNNs explicitly model inter-user spatial correlations; and joint training ensures end-to-end consistency. Contribution/Results: Experiments demonstrate that the proposed method significantly outperforms sub-DFT pilot and iterative precoding baselines under reduced pilot overhead or feedback bit budgets, achieving substantial sum-rate gains. The framework combines high modeling efficiency with strong deployment robustness.

Technology Category

• Machine Learning: Multimodal Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Robust precoding is efficiently feasible in frequency division duplex (FDD) systems by incorporating the learnt statistics of the propagation environment through a generative model. We build on previous work that successfully designed site-specific precoders based on a combination of Gaussian mixture models (GMMs) and graph neural networks (GNNs). In this paper, by utilizing a vector quantized-variational autoencoder (VQ-VAE), we circumvent one of the key drawbacks of GMMs, i.e., the number of GMM components scales exponentially to the feedback bits. In addition, the deep learning architecture of the VQ-VAE allows us to jointly train the GNN together with VQ-VAE along with pilot optimization forming an end-to-end (E2E) model, resulting in considerable performance gains in sum rate for multi-user wireless systems. Simulations demonstrate the superiority of the proposed frameworks over the conventional methods involving the sub-discrete Fourier transform (DFT) pilot matrix and iterative precoder algorithms enabling the deployment of systems characterized by fewer pilots or feedback bits.

Problem

• Designing robust precoders for multi-user FDD wireless systems
• Overcoming exponential scaling of GMM components with feedback bits
• Jointly training GNN and VQ-VAE for end-to-end optimization

Innovation

• VQ-VAE replaces GMM to avoid exponential component scaling
• Jointly trains GNN with VQ-VAE in end-to-end model
• Integrates pilot optimization for reduced feedback bit requirements