ReVeal: A Physics-Informed Neural Network for High-Fidelity Radio Environment Mapping

Le résumé fourni par la source

Accurately mapping the radio environment (e.g., identifying wireless signal strength at specific frequency bands and geographic locations) is crucial for efficient spectrum sharing, enabling Secondary Users (SUs) to access underutilized spectrum bands while protecting Primary Users (PUs). However, existing models either lack generalization due to shadowing, interference, and fading, or are computationally expensive, limiting real-world applicability. To address such shortcomings, we derive a second-order Partial Differential Equation (PDE) for the Received Signal Strength Indicator (RSSI) based on an established statistical model. We then propose ReVeal (Reconstructor and Visualizer of Spectrum Landscape), a novel Physics-Informed Neural Network (PINN) that integrates the PDE residual into a neural network loss function to accurately model the radio environment using sparse Radio Frequency (RF) sensor measurements. ReVeal is validated using real-world measurement data from rural and suburban areas of the ARA testbed and benchmarked against existing methods. ReVeal outperforms traditional approaches in radio environment prediction; for example, with a Root Mean Square Error (RMSE) of only 1.95 dB, ReVeal achieves an accuracy of the order of magnitude higher than existing methods, including 3GPP and ITU-R channel models, ray tracing, and neural networks. In addition, ReVeal achieves high accuracy with low computational complexity while requiring only sparse RF sampling—for instance, just 30 training sample points across a 514-square-kilometer area. The promising results demonstrate ReVeal's potential to advance spectrum management by enabling precise interference management between PUs and SUs.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.