Answers are grounded in this paper and cite their sources.
This paper examines a novel Machine Learning (ML) Radio Frequency (RF) propagation model optimised for urban environments. In defense applications, RF propagation modelling is crucial for estimating the likely performance of both friendly and adversarial RF systems. Traditional empirical RF propagation models are simple and fast but lack accuracy in complex urban settings due to oversimplified assumptions. Physics-based models that solve aspects of Maxwell’s equations are more accurate but require extensive computational resources and precise scene discretization, making them impractical for rapid modelling large complex areas. To address these limitations, an ML-based approach using a Generative Adversarial Network (GAN) is proposed. This ML model functions as an image-to-image translator, converting emitter locations and environmental geometries into RF propagation heat maps. The Pix2Pix GAN architecture, featuring a U-Net generator and PatchGAN discriminator, generates detailed RF maps capturing complex propagation effects like shadowing and channeling. Trained on diverse urban datasets, the model shows an accuracy comparable with physics-based RF propagation models, but with execution speeds that are orders of magnitude faster. The ML model’s efficiency allows for rapid, geometry-specific RF propagation estimates, significantly faster than physics-based models whilst preserving accuracy. This efficiency supports real-time applications, enhancing situational awareness and system optimization in complex environments. In conclusion, the ML-based RF propagation model offers a promising solution for accurate and efficient predictions, leveraging recent advances in generative ML to overcome traditional model limitations and optimize RF system performance in urban scenarios.
No takes yet. Share an insight, caveat, or question.
Catherall et al. (2024) studied this question.