Physics-informed line-of-sight learning for scalable deterministic channel modeling

2026-03-29

AI Summary

This work addresses the high computational cost of line-of-sight (LoS) determination in conventional ray tracing for deterministic channel modeling, which hinders scalability to large-scale scenarios. The authors propose D²LoS, a method that reformulates pixel-level LoS prediction as sparse vertex-level visibility classification coupled with projection point regression, augmented by geometric post-processing to enforce physical constraints. By innovatively integrating physical priors into the neural network architecture, D²LoS effectively suppresses spectral boundary artifacts and enables cross-frequency reuse without retraining. Evaluated on the newly introduced RayVerse-100 urban dataset, D²LoS achieves high accuracy—exhibiting a mean absolute error (MAE) of 3.28 dB in received power, 4.65° in angular spread, and 20.64 ns in delay spread—while accelerating LoS computation by over 25× compared to traditional approaches.

Abstract

Deterministic channel modeling maps a physical environment to its site-specific electromagnetic response. Ray tracing produces complete multi-dimensional channel information but remains prohibitively expensive for area-wide deployment. We identify line-of-sight (LoS) region determination as the dominant bottleneck. To address this, we propose D$^2$LoS, a physics-informed neural network that reformulates dense pixel-level LoS prediction into sparse vertex-level visibility classification and projection point regression, avoiding the spectral bias at sharp boundaries. A geometric post-processing step enforces hard physical constraints, yielding exact piecewise-linear boundaries. Because LoS computation depends only on building geometry, cross-band channel information is obtained by updating material parameters without retraining. We also construct RayVerse-100, a ray-level dataset spanning 100 urban scenarios with per-ray complex gain, angle, delay, and geometric trajectory. Evaluated against rigorous ray tracing ground truth, D$^2$LoS achieves 3.28~dB mean absolute error in received power, 4.65$^\circ$ angular spread error, and 20.64~ns delay spread error, while accelerating visibility computation by over 25$\times$.

Problem

• deterministic channel modeling
• line-of-sight
• ray tracing
• computational bottleneck
• electromagnetic response

Innovation

• Physics-informed neural network
• Line-of-sight prediction
• Deterministic channel modeling
• Geometric post-processing
• Cross-band generalization