Practical Cross-Band Channel Prediction for AI-RAN via Physics-Guided Deep Unfolding

2026-05-29

AI Summary

Existing approaches struggle to simultaneously achieve strong generalization across diverse environments and real-time performance in cross-band channel prediction, limiting their applicability in AI-native radio access networks (AI-RAN). This work proposes GUIDE, a novel framework that introduces physics-guided deep unfolding to this task by embedding differentiable wireless channel physical models directly into neural network layers. This design enables efficient end-to-end prediction without requiring retraining or fine-tuning. The method delivers both high generalization capability and low inference latency in unseen environments, significantly outperforming purely data-driven or model-driven baselines: it achieves 2.75× higher beamforming gain than FIRE and 1.39× higher gain than R2F2, while operating 1,610× faster in inference.

Abstract

To make cross-band channel prediction practical for AI-native RAN, algorithms must generalize across diverse environments and support real-time inference. Existing approaches achieve one but not both. To bridge this gap, we introduce GUIDE, a physics-guided deep unfolding framework that embeds wireless channel physics into differentiable layers. Without retraining in unseen environments, GUIDE achieves 2.75x beamforming gain than the deep learning-based baseline FIRE with only a slight increase in inference time, and 1.39x beamforming gain than the strongest model-based baseline R2F2 while running over 1610x faster.

Problem

• cross-band channel prediction
• AI-RAN
• generalization
• real-time inference
• wireless channel

Innovation

• physics-guided deep unfolding
• cross-band channel prediction
• AI-RAN
• beamforming gain
• real-time inference