Radio Frequency Ray Tracing via Stochastic Geometry

ICLR 2026 Conference Submission23448 Authors

Keywords: Radio Frequency, Neural Rendering, Neural Representation, Stochastic Geometry, Wireless Channel Modeling

Abstract: Radio frequency (RF) propagation modeling is essential for the design, analysis, and optimization of modern wireless sensing and communication systems. However, accurately modeling RF propagation in electrically large and complex environments remains a long-standing challenge, owing to the intricate interactions between RF signals and surrounding objects (e.g., reflection, diffraction, and scattering). Unlike conventional ray-tracing pipelines that hand-engineer interaction rules, or black-box neural surrogates that do not explicitly model physical structure, we introduce RFSG, a novel framework that integrates neural representations with physics-based RF propagation modeling. Starting with a stochastic representation of objects via random indicator functions, we derive the attenuation coefficient as a functional of the probability distributions of the underlying indicator functions under an exponential transport model. This formulation inherently satisfies key physical constraints such as reciprocity and reversibility. Building on this foundation, we employ object-centric neural representations to capture complex RF–object interactions while preserving the composability of traditional ray tracing. Extensive elevations on real-world testbeds demonstrate that RFSG consistently outperforms state-of-the-art neural baselines in prediction accuracy, while requiring significantly fewer training samples.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 23448