ray-tracing / wave-optics simulation

Traditional ray-based models fail to accurately capture wave interference and diffraction, while full-wave simulations suffer from prohibitive computational complexity. To address this, we propose a wave-optical physical modeling method based on bilinear path integrals. This work is the first to extend classical path integral theory to wave optics, establishing a weakly local region-to-region propagation framework. By introducing an elliptical conic geometric structure, we enable efficient path sampling and regional transport, unifying the treatment of interference and diffraction across optical paths. The method significantly improves both accuracy and efficiency in wave-effect simulation. It supports high-fidelity light transport rendering and long-wavelength electromagnetic radiation propagation in complex scenes. As a result, it provides a scalable, physically grounded paradigm for photorealistic rendering and electromagnetic simulation.

To address low ray-tracing efficiency, insufficient multipath modeling accuracy, and manual electromagnetic parameter assignment in point-cloud-driven radio propagation simulation, this paper proposes the first differentiable ray-tracing framework based on raw point clouds. Methodologically, it integrates semantic segmentation labels with differentiable electromagnetic computations to jointly optimize surface electromagnetic parameters—including reflectivity and roughness—enabling physically grounded modeling of specular reflection and diffuse scattering across up to five bounces. Its key innovation lies in achieving end-to-end differentiability directly from unstructured point-cloud input, thereby unifying physical interpretability with data-driven adaptability. Evaluated on two complex indoor scenes, the framework achieves sub-90-ms inference time per simulation while significantly improving multipath modeling accuracy and cross-scene generalization. This work establishes a new paradigm for semantics-aware electromagnetic simulation.

In point-to-point ray tracing, conventional path search suffers from exponential computational complexity and extremely sparse effective paths. Method: This paper proposes a machine learning–driven generative ray-path sampling method—the first to introduce generative modeling into ray-path space. We design a geometrically invariant graph neural network (robust to translation, scaling, and rotation) jointly optimized with differentiable path prioritization, enabling end-to-end trainable dynamic focused sampling. Contribution/Results: Our approach eliminates exhaustive enumeration, reducing computational complexity from exponential to linear. It achieves over 92% path selection accuracy without relying on specific frequency bands or material parameters, and maintains high-fidelity channel modeling across diverse propagation scenarios.

This work addresses the limitations of existing neural scene representations, which prioritize visual appearance and struggle to support deterministic multipath tracing required for radio frequency (RF) propagation, as well as conventional RF simulation methods that rely on manually constructed meshes and lack a unified representation with visual reconstruction. To bridge this gap, the authors propose embedding 3D Gaussian primitives into a hardware-accelerated ray tracing framework, establishing a differentiable, unified model that simultaneously achieves high-quality novel view synthesis and physically plausible RF channel impulse response simulation. Their approach enables, for the first time, direct extraction of multipath RF trajectories from purely vision-driven neural scenes without additional geometric modeling, thereby demonstrating the feasibility and potential of neural representations for RF digital twins.

Modeling rapidly time-varying wireless channels in dynamic scenarios—such as vehicle-to-vehicle (V2V) communications—remains challenging due to complex spatiotemporal propagation effects. Method: This paper systematically compares two mainstream approaches—differentiable ray tracing (DRT) and dynamic ray tracing (Dynamic RT)—and introduces the Multipath Lifetime Map (MPLM), a novel metric that jointly characterizes the spatiotemporal evolution of multipath components solely from static environmental geometry, thereby quantifying channel spatiotemporal coherence. Integrated within the 3DSCAT and Sionna simulation frameworks, the method is validated on a reproducible urban street-canyon scenario. Contribution/Results: Experimental results demonstrate strong agreement between MPLM predictions and measured channel data (mean correlation coefficient > 0.92), establishing MPLM as an interpretable, geometry-driven, and quantitative benchmark for evaluating and selecting dynamic propagation modeling techniques.

This work addresses the limited accuracy of conventional radiomap modeling at high frequencies, which stems from neglecting fine-scale scatterers. The authors propose a multi-scatterer channel model based on spherical wave modal expansion that unifies the characterization of source radiation, single scattering, and multiple-scattering coupling effects through modal superposition. By reformulating the forward model as an inverse optimization problem, the approach jointly estimates scatterer responses and transmitter location. Notably, it integrates multi-scattering interactions with low-order modal approximation within a physically interpretable framework—a first in the field—and enables high-fidelity radiomap reconstruction and extrapolation from sparse measurements. Simulations demonstrate that the proposed model significantly outperforms existing methods in both spatial and beam domains, particularly in dense scattering environments.

This work addresses the high computational complexity of conventional ray tracing methods, which hinders real-time wireless network modeling. The authors propose a novel approach that integrates the physical principles of ray tracing with graph neural networks by converting environmental point clouds into graph structures and leveraging neural message passing to efficiently infer propagation parameters such as signal strength in three-dimensional space. This method establishes the first learnable and generalizable digital twin model of radio environments, enabling joint training on both simulated and real-world measurement data. It achieves high prediction accuracy while significantly reducing inference time, making it well-suited for efficient modeling and prediction in complex 3D wireless scenarios.

This work addresses the limited generalization capability of conventional static ray-tracing methods in complex wireless environments, which often fail to accurately model channel characteristics. To overcome this challenge, the authors propose GAI-NeRF, a novel framework that introduces geometric algebra into neural radiance fields for the first time. The architecture integrates a geometric algebra–based attention mechanism with Transformer-like global token representations to explicitly model electromagnetic interactions between rays and scene objects while aggregating spatial-electromagnetic joint features. Experimental results demonstrate that GAI-NeRF significantly outperforms existing approaches across multiple real-world indoor datasets, achieving notable improvements in both channel prediction accuracy and cross-scenario generalization performance.

This work addresses the challenge of modeling ray-object interactions in complex wireless environments by proposing an end-to-end neural architecture that integrates 3D Gaussian Splatting with a geometric algebra attention mechanism. The model explicitly captures multipath propagation, attenuation, and reflection/diffraction effects of electromagnetic waves. Notably, it introduces geometric algebra into the 3D Gaussian Splatting framework for the first time, enabling joint encoding of spatial geometry and electromagnetic relationships, thereby endowing the model with physical interpretability. Evaluated on multiple real-world indoor datasets, the proposed method achieves state-of-the-art performance across various wireless signal prediction tasks, significantly outperforming existing approaches.

This work addresses the high sensitivity of conventional millimeter-wave ray tracing to holes and noise in geometric reconstructions, which leads to inaccurate channel predictions. The authors propose mmDiff, an end-to-end differentiable framework that, for the first time, integrates a directional scattering model into millimeter-wave ray tracing. By replacing specular reflections with smooth energy distributions, mmDiff significantly enhances robustness to noisy geometry. The method calibrates material parameters directly from sparse measurements via gradient-based optimization and accurately predicts channels in both real-world and synthetic environments. The approach enjoys theoretical guarantees of asymptotic convergence in path gain and demonstrates substantially improved performance over traditional specular-reflection-based methods.