Physics-Informed Neural Networks for Radio Propagation Modeling
Dr. Andreas Molisch, Dr. Dawei Ying
University of Southern California
Abstract
We develop physics-informed neural networks (PINNs) for radio propagation modeling that incorporate Maxwell's equations as soft constraints during training. By encoding electromagnetic wave physics directly into the loss function, our PINNs predict path loss and multipath characteristics with 5x less training data than purely data-driven approaches while maintaining equivalent accuracy. The model generalizes to unseen environments 3x better than standard neural network propagation models.
AI Summary
- Physics-informed neural networks incorporating Maxwell's equations for propagation modeling.
- 5x less training data required compared to purely data-driven approaches.
- 3x better generalization to unseen environments.
- Embeds electromagnetic physics directly into the neural network loss function.
Key Findings
- 1Physics constraints dramatically reduce the amount of measurement data needed for model training.
- 2PINNs correctly predict diffraction effects that data-only models often miss.
- 3The approach works across sub-6 GHz, mmWave, and sub-THz frequency bands.
Industry Implications
Reduces the cost of channel measurement campaigns for 6G network planning.
Enables accurate propagation modeling in new frequency bands with limited data.
Bridges the gap between physics-based and ML-based propagation models.
Read the Original Paper
Access the full paper on arXiv for complete methodology, results, and references.
Open on arXivRelated Papers
Diffusion Models for Wireless Channel Generation and Augmentation
NYU Wireless — 18 citations
AI/ML PapersTransformer-Based Channel Estimation for Massive MIMO Systems
Tsinghua University — 12 citations
AI/ML PapersFederated Reinforcement Learning for Distributed Network Optimization
Stanford University — 8 citations