arxiv: 2605.07781 · v1 · submitted 2026-05-08 · 💻 cs.CV

Recognition: no theorem link

Differentiable Ray Tracing with Gaussians for Unified Radio Propagation Simulation and View Synthesis

Janne Heikkil\"a, Lam Huynh, Miguel Bordallo L\'opez, Niklas Vaara, Pekka Sangi

Authors on Pith no claims yet

Pith reviewed 2026-05-11 02:33 UTC · model grok-4.3

classification 💻 cs.CV

keywords 3D Gaussian SplattingDifferentiable ray tracingRadio propagationChannel impulse responseNeural scene representationView synthesisUnified simulation

0 comments

The pith

Embedding 3D Gaussian Splatting primitives into ray tracing enables joint RF propagation simulation and photorealistic view synthesis from images alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a framework that performs differentiable ray tracing for radio-frequency signals directly inside scenes reconstructed via 3D Gaussian Splatting. Gaussian primitives are placed into a hardware-accelerated ray-tracing structure so that the same representation supports both optical alpha-compositing for novel views and deterministic multi-bounce RF paths with their associated attenuation and delay. A reader would care because conventional RF simulators require hand-crafted meshes while visual reconstruction pipelines ignore electromagnetic properties; a single neural scene could therefore serve as a digital twin for both domains. If the method works, point-to-point channel impulse responses become available at any location inside an image-based model without extra geometry or material calibration steps.

Core claim

By embedding Gaussian primitives into a hardware-accelerated ray tracing structure, the framework computes physically meaningful multi-bounce radio paths and channel impulse responses directly from visual-only 3D Gaussian Splatting reconstructions while preserving real-time, high-fidelity novel view synthesis.

What carries the argument

Embedding of 3D Gaussian Splatting primitives into a hardware-accelerated ray tracing structure that supports both optical compositing and deterministic RF path tracing with attenuation and delay.

If this is right

Point-to-point RF path computation becomes possible between arbitrary 3D locations inside an image-reconstructed scene.
Physically interpretable channel impulse responses can be obtained without building separate meshes for radio simulation.
The same spatial representation supports both differentiable RF simulation and real-time photorealistic rendering.
Cross-modal consistency between visual appearance and electromagnetic propagation is demonstrated on visual-only inputs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Scene reconstruction pipelines could be extended to jointly optimize visual and RF fidelity by back-propagating errors from both domains.
The approach suggests that image-based models may eventually replace manual digital-twin construction for wireless planning in indoor or urban environments.
Material estimation from appearance could be folded into the same differentiable pipeline to reduce reliance on assumed material libraries.

Load-bearing premise

Gaussian primitives placed in the ray tracer accurately encode the scene geometry and material properties needed for correct multi-bounce RF paths without any additional calibration.

What would settle it

Compare the channel impulse responses extracted by the Gaussian ray tracer against ground-truth measurements taken in the same physical scene or against a conventional mesh-based RF simulator on identical geometry.

Figures

Figures reproduced from arXiv: 2605.07781 by Janne Heikkil\"a, Lam Huynh, Miguel Bordallo L\'opez, Niklas Vaara, Pekka Sangi.

**Figure 2.** Figure 2: Corridor scene with NLOS measurements. Measurements and Environment. We compare against the channel measurements conducted in [45]. The measurements consist of a NLoS case, where the frequency was swept in 4001 samples from 110 GHz to 170 GHz in many directional measurements. The geometry was captured with a RGB-D camera. We follow the setup provided in [37], where the measurements are composed into an ag… view at source ↗

**Figure 4.** Figure 4: Max normalized PDPs of NimbusRT v1, Sionna, and our method, compared with the measured one. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 3.** Figure 3: Auditorium scene with one TX and 18 RXs. Measurements and Environment. The channel measurements were carried out in a large auditorium at 234 GHz center frequency. The 4 GHz bandwidth surrounding the center frequency was swept in 1001 uniformly spaced frequency samples. The setup consisted of one TX position and 18 RX positions, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 5.** Figure 5: Unlike Lds, our loss reduces the contribution of the late arriving path to the noise level, which is not observable in the measurements. 0 50 100 150 200 250 Time (ns) 90 80 70 60 Power (dB) Split 6, Test RX 5 Measured Ours ds [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: Our method misses a late arriving propagation path, leading to a substantial underestimation of τrms. Results. The quantitative results are provided in [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

**Figure 7.** Figure 7: RGB image, 2D semantic labels, and rendered labels with naive and alpha weighted assignment. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗

**Figure 8.** Figure 8: Max normalized power delay profiles of the ablation cases. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗

**Figure 9.** Figure 9: Ground truth RGB image, as well as the rendered RGB, normal, depth and edges by our method in the [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗

**Figure 10.** Figure 10: Label post-processing [PITH_FULL_IMAGE:figures/full_fig_p018_10.png] view at source ↗

**Figure 11.** Figure 11: Ground truth RGB image, as well as the rendered RGB, normal, depth and edges by our method in [PITH_FULL_IMAGE:figures/full_fig_p018_11.png] view at source ↗

**Figure 12.** Figure 12: Qualitative results for all test RXs splits. The simulated ones are with applied noise computed from [PITH_FULL_IMAGE:figures/full_fig_p019_12.png] view at source ↗

**Figure 13.** Figure 13: Novel views rendered with different methods. Mip-NeRF360 and 3DGS results were taken from [ [PITH_FULL_IMAGE:figures/full_fig_p021_13.png] view at source ↗

**Figure 14.** Figure 14: Received signal strength predictions at 1084 receiver locations provided in the RF3DGS [ [PITH_FULL_IMAGE:figures/full_fig_p022_14.png] view at source ↗

**Figure 15.** Figure 15: Ground truth RGB image, as well as the rendered RGB, normal, depth and edges by our method in [PITH_FULL_IMAGE:figures/full_fig_p022_15.png] view at source ↗

**Figure 16.** Figure 16: Measurement setup in the scene and loss graph for the experiments in this environment. [PITH_FULL_IMAGE:figures/full_fig_p023_16.png] view at source ↗

**Figure 17.** Figure 17: Ground truth RGB image, as well as the rendered RGB, normal, depth and edges by our method in [PITH_FULL_IMAGE:figures/full_fig_p024_17.png] view at source ↗

**Figure 18.** Figure 18: Ray traced paths with and without edge-aware ray tracing. [PITH_FULL_IMAGE:figures/full_fig_p024_18.png] view at source ↗

read the original abstract

Explicit neural representations such as 3D Gaussian Splatting (3DGS) enable high-fidelity and real-time novel view synthesis, yet optimize for alpha-composited optical appearance rather than ray-intersectable geometry. In contrast, radio-frequency (RF) digital twins require deterministic multi-bounce paths, where the geometry dictates trajectories and their associated attenuation and delay. We introduce a framework enabling differentiable RF propagation simulation directly within visually reconstructed neural scenes, allowing point-to-point path computation between arbitrary 3D locations while preserving high-quality visual rendering. Unlike conventional RF simulation pipelines that rely on manually constructed meshes, we embed Gaussian primitives into a hardware-accelerated ray tracing structure as the underlying spatial representation. By extracting physically meaningful channel impulse responses from visual-only reconstructions, we provide cross-modal evidence that neural reconstructions can serve as unified spatial representations for both electromagnetic propagation simulation and photorealistic view synthesis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper embeds 3D Gaussians into a hardware ray tracer so RF paths can be computed directly from a visually trained scene, but the physical fidelity of those paths is the part that still needs checking.

read the letter

The main thing to know is that they load Gaussian primitives into a hardware-accelerated ray-tracing structure so you can trace deterministic multi-bounce RF paths inside a 3DGS scene that was built only from images. This gives you channel impulse responses without first converting the scene to a mesh, which is the practical step forward over standard RF pipelines. The unification is real: the same representation supports both photorealistic view synthesis and point-to-point radio simulation, and the abstract shows they keep the visual quality while adding the RF capability. That is the concrete new piece. They do a clean job describing how the embedding works and why it avoids the manual geometry step that usually separates the two domains. For anyone building digital twins or integrated sensing setups, that shortcut matters. The soft spot is the one the stress test flags. Gaussians optimized for alpha-composited RGB appearance produce soft ellipsoids rather than hard surfaces with precise normals and dielectric boundaries. The paper does not appear to add a mechanism for assigning or inferring RF material parameters such as permittivity or conductivity from visual data alone, so the extracted delays and attenuations rest on the assumption that visual geometry is already good enough for electromagnetic accuracy. If the full results include direct comparisons against measured RF data or against mesh-based simulators, that would address the concern; without it the claim of physically meaningful outputs is still provisional. This is for readers working at the intersection of neural rendering and wireless simulation who want a single scene model for both tasks. It is not yet a finished tool for production RF work, but the framework is clear enough that a serious referee should see it. I would send it out for review so the authors can supply the validation numbers and clarify how the ray intersections are handled for reflection and attenuation.

Referee Report

2 major / 2 minor

Summary. The paper introduces a framework that embeds 3D Gaussian Splatting primitives into a hardware-accelerated ray-tracing structure, enabling differentiable computation of multi-bounce RF paths (including delay and attenuation) directly inside scenes reconstructed from RGB images alone. It claims this yields physically meaningful channel impulse responses while preserving real-time photorealistic view synthesis, thereby providing a unified spatial representation for both electromagnetic propagation and visual rendering without requiring manually constructed meshes.

Significance. If validated, the approach would reduce the cost of building RF digital twins by reusing visual reconstructions, offering cross-modal evidence that neural scene representations can support deterministic physics-based simulation. The hardware ray-tracing integration and differentiability are practical strengths; however, the significance is currently limited by the absence of demonstrated physical fidelity for material-dependent effects.

major comments (2)

[Abstract and §3] Abstract and §3 (method overview): the central claim that 'physically meaningful channel impulse responses' are extracted from visual-only reconstructions is load-bearing, yet the description provides no mechanism for assigning or inferring RF material parameters (permittivity, conductivity, roughness) from RGB-optimized Gaussians; without this, extracted delays and attenuations risk being artifacts of the visual loss rather than electromagnetic simulation.
[§4] §4 (ray-tracing embedding): the use of soft, alpha-blended Gaussian ellipsoids for intersection and normal computation is not shown to produce the hard dielectric boundaries required for correct reflection coefficients and multi-bounce paths; the paper must demonstrate that the resulting geometry yields deterministic, physically accurate trajectories rather than approximate splat intersections.

minor comments (2)

[Figure 2 and §5] Figure 2 and §5 (experiments): add quantitative comparison of simulated CIRs against ground-truth measurements or ray-tracing on explicit meshes to substantiate the 'physically meaningful' claim.
[§3.2] Notation in §3.2: define the mapping from Gaussian covariance to surface normal and intersection point explicitly, including any approximation used for hardware acceleration.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment point by point below, clarifying our approach and indicating planned revisions to improve precision and validation.

read point-by-point responses

Referee: [Abstract and §3] Abstract and §3 (method overview): the central claim that 'physically meaningful channel impulse responses' are extracted from visual-only reconstructions is load-bearing, yet the description provides no mechanism for assigning or inferring RF material parameters (permittivity, conductivity, roughness) from RGB-optimized Gaussians; without this, extracted delays and attenuations risk being artifacts of the visual loss rather than electromagnetic simulation.

Authors: We agree that the manuscript requires greater clarity on this distinction. Our framework computes deterministic multi-bounce paths and delays directly from the explicit 3D Gaussian geometry via hardware ray tracing; attenuations and reflection coefficients are then derived using user-specified or default material parameters applied to the intersected surfaces. No mechanism for inferring permittivity, conductivity, or roughness from RGB-optimized Gaussians is present, as the visual reconstruction does not encode electromagnetic properties. The term 'physically meaningful' refers specifically to the use of ray-traced trajectories on the reconstructed scene rather than implicit or learned propagation models. We will revise the abstract and §3 to explicitly state these assumptions, note that material assignment is orthogonal to the geometric representation, and emphasize that path computation is independent of the visual optimization loss after the initial reconstruction. revision: partial
Referee: [§4] §4 (ray-tracing embedding): the use of soft, alpha-blended Gaussian ellipsoids for intersection and normal computation is not shown to produce the hard dielectric boundaries required for correct reflection coefficients and multi-bounce paths; the paper must demonstrate that the resulting geometry yields deterministic, physically accurate trajectories rather than approximate splat intersections.

Authors: We acknowledge the need for explicit validation of the intersection model. Although alpha blending is used only for the differentiable visual renderer, the RF simulation treats each Gaussian as a hard ellipsoid surface for ray intersection tests within the hardware-accelerated structure, yielding deterministic hit points and surface normals for reflection coefficient computation. To demonstrate physical accuracy, we will add quantitative comparisons in the revised §4 and experiments section: path trajectories, bounce counts, and resulting channel impulse responses will be evaluated against reference simulations performed on manually constructed meshes of the same environments. Visualizations of intersection points and normals will also be included to confirm the hard-boundary behavior. revision: yes

Circularity Check

0 steps flagged

No circularity: new embedding framework is independently proposed

full rationale

The paper's derivation chain consists of introducing a novel embedding of 3D Gaussian primitives into a hardware-accelerated ray-tracing structure, then using that structure to compute point-to-point RF paths and extract CIRs. This is presented as an extension of existing 3DGS (cited externally) rather than a reduction of any output to fitted inputs or self-referential definitions. No equation or claim reduces a 'prediction' to a parameter fit by construction, nor does any load-bearing step rely on a self-citation chain that itself assumes the target result. The central claim of cross-modal evidence is grounded in the proposed unified representation and its implementation, making the work self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests primarily on a domain assumption about the transferability of Gaussian representations to RF physics; no free parameters or invented entities are identifiable from the provided text.

axioms (1)

domain assumption Gaussian primitives embedded in ray tracing structures can represent geometry sufficiently for accurate multi-bounce RF propagation including attenuation and delay
Invoked when stating that the framework extracts physically meaningful channel impulse responses from visual reconstructions.

pith-pipeline@v0.9.0 · 5467 in / 1327 out tokens · 75028 ms · 2026-05-11T02:33:59.758353+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages · 3 internal anchors

[1]

3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42(4), 2023

Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42(4), 2023

work page 2023
[2]

2d gaussian splatting for geometrically accurate radiance fields

Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically accurate radiance fields. InACM SIGGRAPH 2024 Conference Papers, pages 1–11, 2024

work page 2024
[3]

Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction.IEEE Transactions on Visualization and Computer Graphics, 31(9):6100–6111, 2024

Danpeng Chen, Hai Li, Weicai Ye, Yifan Wang, Weijian Xie, Shangjin Zhai, Nan Wang, Haomin Liu, Hujun Bao, and Guofeng Zhang. Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction.IEEE Transactions on Visualization and Computer Graphics, 31(9):6100–6111, 2024

work page 2024
[4]

Dn-splatter: Depth and normal priors for gaussian splatting and meshing

Matias Turkulainen, Xuqian Ren, Iaroslav Melekhov, Otto Seiskari, Esa Rahtu, and Juho Kannala. Dn-splatter: Depth and normal priors for gaussian splatting and meshing. In2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 2421–2431. IEEE, 2025

work page 2025
[5]

Accessed: 01.03.2026

ITU-R P.2040.https://www.itu.int/rec/R-REC-P.2040/en. Accessed: 01.03.2026

work page 2040
[6]

Learning radio environments by differentiable ray tracing.IEEE Transactions on Machine Learning in Communications and Networking, 2: 1527–1539, 2024

Jakob Hoydis, Fayçal Aït Aoudia, Sebastian Cammerer, Florian Euchner, Merlin Nimier-David, Stephan Ten Brink, and Alexander Keller. Learning radio environments by differentiable ray tracing.IEEE Transactions on Machine Learning in Communications and Networking, 2: 1527–1539, 2024

work page 2024
[7]

Learnable wireless digital twins: Reconstructing electromagnetic field with neural representations.IEEE Open Journal of the Communications Society, 6:1568–1590, 2025

Shuaifeng Jiang, Qi Qu, Xiaqing Pan, Abhishek K Agrawal, Richard Newcombe, and Ahmed Alkhateeb. Learnable wireless digital twins: Reconstructing electromagnetic field with neural representations.IEEE Open Journal of the Communications Society, 6:1568–1590, 2025

work page 2025
[8]

Sionna RT: Technical report,

Fayçal Aït Aoudia, Jakob Hoydis, Merlin Nimier-David, Baptiste Nicolet, Sebastian Cammerer, and Alexander Keller. Sionna rt: Technical report.arXiv preprint arXiv:2504.21719, 2025

work page arXiv 2025
[9]

Nerf: Representing scenes as neural radiance fields for view synthesis

Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoor- thi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1):99–106, 2021

work page 2021
[10]

Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering

Antoine Guédon and Vincent Lepetit. Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5354–5363, 2024

work page 2024
[11]

Gaussianshader: 3d gaussian splatting with shading functions for reflective surfaces

Yingwenqi Jiang, Jiadong Tu, Yuan Liu, Xifeng Gao, Xiaoxiao Long, Wenping Wang, and Yuexin Ma. Gaussianshader: 3d gaussian splatting with shading functions for reflective surfaces. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5322–5332, 2024

work page 2024
[12]

Relightable 3d gaussians: Realistic point cloud relighting with brdf decomposition and ray tracing

Jian Gao, Chun Gu, Youtian Lin, Zhihao Li, Hao Zhu, Xun Cao, Li Zhang, and Yao Yao. Relightable 3d gaussians: Realistic point cloud relighting with brdf decomposition and ray tracing. InEuropean Conference on Computer Vision, pages 73–89. Springer, 2024. 10

work page 2024
[13]

arXiv preprint arXiv:2412.15867 , year=

Chun Gu, Xiaofei Wei, Zixuan Zeng, Yuxuan Yao, and Li Zhang. Irgs: Inter-reflective gaussian splatting with 2d gaussian ray tracing.arXiv preprint arXiv:2412.15867, 2024

work page arXiv 2024
[14]

Spectre-gs: Modeling highly specular surfaces with reflected nearby objects by tracing rays in 3d gaussian splatting

Jiajun Tang, Fan Fei, Zhihao Li, Xiao Tang, Shiyong Liu, Youyu Chen, Binxiao Huang, Zhenyu Chen, Xiaofei Wu, and Boxin Shi. Spectre-gs: Modeling highly specular surfaces with reflected nearby objects by tracing rays in 3d gaussian splatting. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 16133–16142, 2025

work page 2025
[15]

3d gaussian ray tracing: Fast tracing of particle scenes.ACM Transactions on Graphics (TOG), 43(6):1–19, 2024

Nicolas Moenne-Loccoz, Ashkan Mirzaei, Or Perel, Riccardo de Lutio, Janick Martinez Esturo, Gavriel State, Sanja Fidler, Nicholas Sharp, and Zan Gojcic. 3d gaussian ray tracing: Fast tracing of particle scenes.ACM Transactions on Graphics (TOG), 43(6):1–19, 2024

work page 2024
[16]

Envgs: Modeling view-dependent appearance with environment gaussian

Tao Xie, Xi Chen, Zhen Xu, Yiman Xie, Yudong Jin, Yujun Shen, Sida Peng, Hujun Bao, and Xiaowei Zhou. Envgs: Modeling view-dependent appearance with environment gaussian. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 5742–5751, 2025

work page 2025
[17]

Winert: Towards neural ray tracing for wireless channel modelling and differentiable simulations

Tribhuvanesh Orekondy, Pratik Kumar, Shreya Kadambi, Hao Ye, Joseph Soriaga, and Arash Behboodi. Winert: Towards neural ray tracing for wireless channel modelling and differentiable simulations. InThe Eleventh International Conference on Learning Representations, 2023

work page 2023
[18]

Nerf2: Neural radio-frequency radiance fields

Xiaopeng Zhao, Zhenlin An, Qingrui Pan, and Lei Yang. Nerf2: Neural radio-frequency radiance fields. InProceedings of the 29th Annual International Conference on Mobile Computing and Networking, pages 1–15, 2023

work page 2023
[19]

Newrf: a deep learning framework for wireless radiation field reconstruction and channel prediction

Haofan Lu, Christopher Vattheuer, Baharan Mirzasoleiman, and Omid Abari. Newrf: a deep learning framework for wireless radiation field reconstruction and channel prediction. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org, 2024

work page 2024
[20]

Wrf-gs: Wireless radiation field reconstruction with 3d gaussian splatting

Chaozheng Wen, Jingwen Tong, Yingdong Hu, Zehong Lin, and Jun Zhang. Wrf-gs: Wireless radiation field reconstruction with 3d gaussian splatting. InIEEE INFOCOM 2025-IEEE Conference on Computer Communications, pages 1–10. IEEE, 2025

work page 2025
[21]

Neural repre- sentation for wireless radiation field reconstruction: A 3d gaussian splatting approach.IEEE Transactions on Wireless Communications, 2025

Chaozheng Wen, Jingwen Tong, Yingdong Hu, Zehong Lin, and Jun Zhang. Neural repre- sentation for wireless radiation field reconstruction: A 3d gaussian splatting approach.IEEE Transactions on Wireless Communications, 2025

work page 2025
[22]

Gsrf: Complex-valued 3d gaussian splatting for efficient radio-frequency data synthesis.arXiv preprint arXiv:2502.01826, 2025

Kang Yang, Gaofeng Dong, Sijie Ji, Wan Du, and Mani Srivastava. Gsrf: Complex-valued 3d gaussian splatting for efficient radio-frequency data synthesis.arXiv preprint arXiv:2502.01826, 2025

work page arXiv 2025
[23]

Rf- 3dgs: Wireless channel modeling with radio radiance field and 3d gaussian splatting.IEEE Transactions on Wireless Communications, 25:10419–10433, 2026

Lihao Zhang, Haijian Sun, Samuel Berweger, Camillo Gentile, and Rose Qingyang Hu. Rf- 3dgs: Wireless channel modeling with radio radiance field and 3d gaussian splatting.IEEE Transactions on Wireless Communications, 25:10419–10433, 2026

work page 2026
[24]

Simplification method of 3d point cloud data for ray trace simulation in indoor environment.IEICE Communications Express, 9(6):182–187, 2020

Wataru Okamura, Rikisenia Lukita, Gilbert Ching, Yuki Matsuyama, Yukiko Kishiki, Zhihang Chen, Kentaro Saito, and Jun-ichi Takada. Simplification method of 3d point cloud data for ray trace simulation in indoor environment.IEICE Communications Express, 9(6):182–187, 2020

work page 2020
[25]

A gpu-based radio wave propagation prediction with progressive processing on point cloud.IEEE Antennas and Wireless Propagation Letters, 20(6):1078–1082, 2021

Mingjie Pang, Han Wang, Kaiwei Lin, and Hai Lin. A gpu-based radio wave propagation prediction with progressive processing on point cloud.IEEE Antennas and Wireless Propagation Letters, 20(6):1078–1082, 2021

work page 2021
[26]

From 3d point cloud data to ray-tracing multi-band simulations in industrial scenario

Han Niu, Diego Dupleich, Yanneck Völker-Schöneberg, Alexander Ebert, Robert Müller, Joseph Eichinger, Alexander Artemenko, Giovanni Del Galdo, and Reiner S Thomä. From 3d point cloud data to ray-tracing multi-band simulations in industrial scenario. In2022 IEEE 95th V ehicular Technology Conference:(VTC2022-Spring), pages 1–5. IEEE, 2022

work page 2022
[27]

Indoor radio map construction via ray tracing with rgb-d sensor-based 3d reconstruction: Concept and experiments in wlan systems.IEEE Access, 11:24863–24874, 2023

Norisato Suga, Yoshihiro Maeda, and Koya Sato. Indoor radio map construction via ray tracing with rgb-d sensor-based 3d reconstruction: Concept and experiments in wlan systems.IEEE Access, 11:24863–24874, 2023. 11

work page 2023
[28]

Environment- driven mmwave beamforming for multi-user immersive applications

Ahmad Kamari, Nan Wu, Hemant Kumar Narsani, Bo Han, and Parth Pathak. Environment- driven mmwave beamforming for multi-user immersive applications. InProceedings of the 1st ACM Workshop on Mobile Immersive Computing, Networking, and Systems, pages 268–275, 2023

work page 2023
[29]

Guozhen Xia, Chen Zhou, Fubin Zhang, Zeyu Cui, Chengkai Liu, Hao Ji, Xinmiao Zhang, Zhengyu Zhao, and Yin Xiao. Path loss prediction in urban environments with sionna-rt based on accurate propagation scene models at 2.8 ghz.IEEE Transactions on Antennas and Propagation, 72(10):7986–7997, 2024

work page 2024
[30]

Rgb-d sensor-aided radio map estimation using materials classification.IEEE Wireless Communica- tions Letters, 2025

Norisato Suga, Naoya Yoshida, Ryotaro Gozono, Yoshihiro Maeda, and Koya Sato. Rgb-d sensor-aided radio map estimation using materials classification.IEEE Wireless Communica- tions Letters, 2025

work page 2025
[31]

Sionna rt: Differentiable ray tracing for radio propagation modeling

Jakob Hoydis, Fayçal Aït Aoudia, Sebastian Cammerer, Merlin Nimier-David, Nikolaus Binder, Guillermo Marcus, and Alexander Keller. Sionna rt: Differentiable ray tracing for radio propagation modeling. In2023 IEEE Globecom Workshops (GC Wkshps), pages 317–321. IEEE, 2023

work page 2023
[32]

https://www.remcom.com/wireless-insite-em-propagation-software

Wireless Insite. https://www.remcom.com/wireless-insite-em-propagation-software . Accessed: 01.03.2026

work page 2026
[33]

Simulating specular reflections for point cloud geometrical database of the environment

Usman Tahir Virk, Jean-Frederic Wagen, and Katsuyuki Haneda. Simulating specular reflections for point cloud geometrical database of the environment. In2015 Loughborough Antennas & Propagation Conference (LAPC), pages 1–5. IEEE, 2015

work page 2015
[34]

Indoor propagation channel simulations at 60 ghz using point cloud data.IEEE Transactions on Antennas and Propagation, 64(10): 4457–4467, 2016

Jan Järveläinen, Katsuyuki Haneda, and Aki Karttunen. Indoor propagation channel simulations at 60 ghz using point cloud data.IEEE Transactions on Antennas and Propagation, 64(10): 4457–4467, 2016

work page 2016
[35]

Point cloud ray-launching simulations of indoor multipath channels at 60 ghz

Pasi Koivumäki and Katsuyuki Haneda. Point cloud ray-launching simulations of indoor multipath channels at 60 ghz. In2022 IEEE 33rd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), pages 01–07. IEEE, 2022

work page 2022
[36]

Ray-optics simulations of outdoor-to- indoor multipath channels at 4 and 14 ghz.IEEE Transactions on Antennas and Propagation, 71(7):6046–6059, 2023

Pasi Koivumäki, Aki Karttunen, and Katsuyuki Haneda. Ray-optics simulations of outdoor-to- indoor multipath channels at 4 and 14 ghz.IEEE Transactions on Antennas and Propagation, 71(7):6046–6059, 2023

work page 2023
[37]

Ray launching-based computation of exact paths with noisy dense point clouds.IEEE Transactions on Antennas and Propagation, 2025

Niklas Vaara, Pekka Sangi, Miguel Bordallo López, and Janne Heikkilä. Ray launching-based computation of exact paths with noisy dense point clouds.IEEE Transactions on Antennas and Propagation, 2025

work page 2025
[38]

Differentiable high- performance ray tracing-based simulation of radio propagation with point clouds.IEEE Anten- nas and Wireless Propagation Letters, pages 1–5, 2026

Niklas Vaara, Pekka Sangi, Miguel Bordallo López, and Janne Heikkilä. Differentiable high- performance ray tracing-based simulation of radio propagation with point clouds.IEEE Anten- nas and Wireless Propagation Letters, pages 1–5, 2026. doi: 10.1109/LAWP.2026.3670638

work page doi:10.1109/lawp.2026.3670638 2026
[39]

Radio propagation modelling: To differentiate or to deep learn, that is the question,

Stefanos Bakirtzis, Paul Almasan, José Suárez-Varela, Gabriel O Ferreira, Michail Kalntis, An- dré Felipe Zanella, Ian Wassell, and Andra Lutu. Radio propagation modelling: To differentiate or to deep learn, that is the question.arXiv preprint arXiv:2509.19337, 2025

work page arXiv 2025
[40]

Oriane Siméoni, Huy V V o, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Ramamonjisoa, et al. Dinov3. arXiv preprint arXiv:2508.10104, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[41]

Scene parsing through ade20k dataset

Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene parsing through ade20k dataset. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 633–641, 2017

work page 2017
[42]

Gaussian opacity fields: Efficient and compact surface reconstruction in unbounded scenes.arXiv preprint arXiv:2404.10772, 2024

Zehao Yu, Torsten Sattler, and Andreas Geiger. Gaussian opacity fields: Efficient and compact surface reconstruction in unbounded scenes.arXiv preprint arXiv:2404.10772, 2024

work page arXiv 2024
[43]

MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details

Ruicheng Wang, Sicheng Xu, Yue Dong, Yu Deng, Jianfeng Xiang, Zelong Lv, Guangzhong Sun, Xin Tong, and Jiaolong Yang. Moge-2: Accurate monocular geometry with metric scale and sharp details.arXiv preprint arXiv:2507.02546, 2025. 12

work page internal anchor Pith review arXiv 2025
[44]

Accessed: 01.03.2026

slang-torch.https://github.com/shader-slang/slang-torch. Accessed: 01.03.2026

work page 2026
[45]

Initial results on d band channel measurements in los and nlos office corridor environment

Joonas Kokkoniemi, Veikko Hovinen, Klaus Nevala, and Markku Juntti. Initial results on d band channel measurements in los and nlos office corridor environment. In2022 16th European Conference on Antennas and Propagation (EuCAP), pages 1–5. IEEE, 2022

work page 2022
[46]

Absgs: Recovering fine details in 3d gaussian splatting

Zongxin Ye, Wenyu Li, Sidun Liu, Peng Qiao, and Yong Dou. Absgs: Recovering fine details in 3d gaussian splatting. InProceedings of the 32nd ACM International Conference on Multimedia, pages 1053–1061, 2024

work page 2024
[47]

Citygaussianv2: Efficient and geometrically accurate reconstruction for large-scale scenes

Yang Liu, Chuanchen Luo, Zhongkai Mao, Junran Peng, and Zhaoxiang Zhang. Citygaussianv2: Efficient and geometrically accurate reconstruction for large-scale scenes. InThe Thirteenth International Conference on Learning Representations, 2025. URL https://openreview. net/forum?id=a3ptUbuzbW

work page 2025
[48]

Stopthepop: Sorted gaussian splatting for view-consistent real-time rendering.ACM Transactions on Graphics (TOG), 43(4):1–17, 2024

Lukas Radl, Michael Steiner, Mathias Parger, Alexander Weinrauch, Bernhard Kerbl, and Markus Steinberger. Stopthepop: Sorted gaussian splatting for view-consistent real-time rendering.ACM Transactions on Graphics (TOG), 43(4):1–17, 2024

work page 2024
[49]

Bs3d: Building-scale 3d reconstruction from rgb-d images

Janne Mustaniemi, Juho Kannala, Esa Rahtu, Li Liu, and Janne Heikkilä. Bs3d: Building-scale 3d reconstruction from rgb-d images. InScandinavian Conference on Image Analysis, pages 551–565. Springer, 2023

work page 2023
[50]

Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields

Jonathan T Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields. InProceedings of the IEEE/CVF international conference on computer vision, pages 5855–5864, 2021

work page 2021
[51]

Deep blending for free-viewpoint image-based rendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018

Peter Hedman, Julien Philip, True Price, Jan-Michael Frahm, George Drettakis, and Gabriel Brostow. Deep blending for free-viewpoint image-based rendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018

work page 2018
[52]

Tanks and temples: Bench- marking large-scale scene reconstruction.ACM Transactions on Graphics (ToG), 36(4):1–13, 2017

Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: Bench- marking large-scale scene reconstruction.ACM Transactions on Graphics (ToG), 36(4):1–13, 2017

work page 2017
[53]

Pixelwise view selection for unstructured multi-view stereo

Johannes Lutz Schönberger, Enliang Zheng, Marc Pollefeys, and Jan-Michael Frahm. Pixelwise view selection for unstructured multi-view stereo. InEuropean Conference on Computer Vision (ECCV), 2016

work page 2016
[54]

Structure-from-motion revisited

Johannes Lutz Schönberger and Jan-Michael Frahm. Structure-from-motion revisited. In Conference on Computer Vision and Pattern Recognition (CVPR), 2016

work page 2016
[55]

S. Umeyama. Least-squares estimation of transformation parameters between two point patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence, 13(4):376–380, 1991. doi: 10.1109/34.88573

work page doi:10.1109/34.88573 1991
[56]

Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J. Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, Anton Clarkson, Mingfei Yan, Brian Budge, Yajie Yan, Xiaqing Pan, June Yon, Yuyang Zou, Kimberly Leon, Nigel Carter, Jesus Briales, Tyler Gillingham, Elias Mueggler, Luis Pesqueira, Manolis Savva, Dhruv Batra, Hauke M. S...

work page internal anchor Pith review arXiv 1906