Robust Super-Capacity SRS Channel Inpainting via Diffusion Models

2025-10-29

AI Summary

To address the low CSI recovery accuracy caused by sparse and non-uniform Sounding Reference Signal (SRS) allocation in 5G NR, and the poor generalization of existing Masked Autoencoder (MAE)-based methods—prone to overfitting training masks and failing under unseen distortions such as interference, clipping, and non-Gaussian noise—this paper proposes a robust channel reconstruction framework based on score-based diffusion models. Our method innovatively incorporates a system model prior, dynamically guiding denoising inference via a likelihood gradient term, thereby significantly enhancing adaptability to distribution shifts. We adopt a U-Net architecture conditioned on masked SRS inputs and employ a gradient-enhanced iterative sampling strategy. Experiments on Channel Distribution Models (CDL) demonstrate that our approach achieves up to 14 dB improvement in normalized mean square error (NMSE) over MAE and U-Net baselines, while maintaining competitive performance in matched scenarios—marking the first demonstration of a single diffusion model enabling highly robust CSI reconstruction across multiple distortion types.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Accurate channel state information (CSI) is essential for reliable multiuser MIMO operation. In 5G NR, reciprocity-based beamforming via uplink Sounding Reference Signals (SRS) face resource and coverage constraints, motivating sparse non-uniform SRS allocation. Prior masked-autoencoder (MAE) approaches improve coverage but overfit to training masks and degrade under unseen distortions (e.g., additional masking, interference, clipping, non-Gaussian noise). We propose a diffusion-based channel inpainting framework that integrates system-model knowledge at inference via a likelihood-gradient term, enabling a single trained model to adapt across mismatched conditions. On standardized CDL channels, the score-based diffusion variant consistently outperforms a UNet score-model baseline and the one-step MAE under distribution shift, with improvements up to 14 dB NMSE in challenging settings (e.g., Laplace noise, user interference), while retaining competitive accuracy under matched conditions. These results demonstrate that diffusion-guided inpainting is a robust and generalizable approach for super-capacity SRS design in 5G NR systems.

Problem

• Improving CSI accuracy for 5G MIMO systems
• Addressing SRS sparsity and coverage constraints
• Handling unseen distortions and distribution shifts

Innovation

• Diffusion-based framework for channel inpainting
• Integrates system-model knowledge via likelihood-gradient
• Outperforms baseline models under distribution shift