arxiv: 2604.07086 · v1 · submitted 2026-04-08 · 📡 eess.SP

Recognition: unknown

Radio-Frequency Inverse Rendering for Wireless Environment Modeling

Fuhai Wang , Zihan Jin , Lehang Wang , Xuehui Dong , Tiebin Mi , Robert Caiming Qiu , Zenan ling

Authors on Pith no claims yet

Pith reviewed 2026-05-10 17:11 UTC · model grok-4.3

classification 📡 eess.SP

keywords radio frequencyinverse renderingGaussian splattingwireless modelingradar cross-sectionsignal strength predictionscene editing

0 comments

The pith

A radio-frequency rendering method decouples emission, geometry, and materials to support editable wireless models and multiple prediction tasks.

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

The paper presents a framework that treats radio-frequency scenes like visual scenes in rendering but keeps the physical emission, surface shapes, and material responses separate. Prior approaches mixed these elements together, which made it hard to change the scene or adjust the wireless setup without rebuilding the whole model. By placing an RF-specific scattering function inside a Gaussian splatting representation and tracing rays through it, the method can handle three different radio tasks with the same underlying structure. Experiments show clear gains on each task compared with entangled baselines. If the separation works reliably, it opens the door to treating wireless environments as editable, queryable models rather than fixed black boxes.

Core claim

The central claim is that an RF inverse rendering framework, built by embedding an RF-aware bidirectional scattering distribution function into the Gaussian splatting paradigm as an RF rendering equation, can explicitly separate RF emission from geometry and material electromagnetic properties; each Gaussian primitive carries surface normals, material parameters, and roughness, which a customized ray-tracing scheme then uses to synthesize RF signals, allowing the same model to perform radar cross-section synthesis, received signal strength indicator prediction, and wireless scene editing.

What carries the argument

An RF-aware bidirectional scattering distribution function placed inside Gaussian splatting and driven by a customized ray-tracing scheme.

If this is right

The same trained representation can generate radar cross-section patterns for objects by varying incident angles and materials without retraining.
Received signal strength at any receiver location can be computed directly from the decoupled geometry and material attributes.
Changing the position, shape, or material of any surface in the scene updates all downstream RF predictions without rebuilding the model.
Wireless system parameters such as transmitter placement or antenna orientation can be optimized by querying the explicit physical attributes.

Where Pith is reading between the lines

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

The decoupled structure could support hybrid models that combine radio data with visible-light images for joint environment reconstruction.
Real-time updates to the Gaussian primitives might enable dynamic wireless scene modeling for moving objects or changing materials.
The same separation of concerns could be tested on other wave phenomena such as acoustic propagation or light transport in participating media.

Load-bearing premise

The assumption that giving Gaussian primitives physical surface normals, electromagnetic material parameters, and roughness values, then tracing rays through them, will reproduce real radio propagation without large systematic errors.

What would settle it

A side-by-side comparison of the model's predicted received signal strength values against calibrated measurements taken in a known indoor room with documented wall materials and exact transmitter-receiver locations.

Figures

Figures reproduced from arXiv: 2604.07086 by Fuhai Wang, Lehang Wang, Robert Caiming Qiu, Tiebin Mi, Xuehui Dong, Zenan ling, Zihan Jin.

**Figure 1.** Figure 1: RF Inverse Rendering. (a) Classical RF rendering and (b) the proposed RF inverse rendering. Inverse rendering enables decoupled RF attribute rendering (c) and broader applications (d). 3. Preliminaries This section reviews NeRF and 3DGS as neural volumetric representations for wireless channel modeling. NeRF-based Channel Modeling. To model radio radiance fields, NeRF-based methods (Zhao et al., 2023; Luo… view at source ↗

**Figure 2.** Figure 2: Overview of system workflow for each Gaussian primitive, and an optimization strategy tailored for RF-aware physically based rendering is devised, as detailed in the Appendix E. Specifically, the hybrid logic in (1) is extended to synthesize the scene’s depth map D and normal map N as follows: {D, N } = X i∈N αiTi{di , ni} where di denotes the z-depth coordinate of the i-th Gaussian in the view space. By … view at source ↗

**Figure 4.** Figure 4: Frequency-Aware Attribute Modulation Network (FAM). We present a novel approach that employs attribute control together with a deformation MLP to modulate the frequency-dependent behavior of 3D Gaussians. The transformed primitives are then rendered into broadband radio signals. During the RF rendering stage, the RF-aware attributes {α, R, |Γ|, ∠Γ} are optimized through inverse rendering of the observed RF… view at source ↗

**Figure 6.** Figure 6: The reconstructed wideband RCS results. accomplished via a deformable RF attributes MLP, which adaptively modulates the Gaussian attributes to capture frequency-dependent variations. The prediction results are shown in [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

**Figure 8.** Figure 8: Predicted RCS of the target at 3 m and 2.4 GHz. m, 4.5 m, and 5 m, and test on unseen data at 3 m. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

**Figure 7.** Figure 7: Predicted radio map of the classroom. characterize wireless scattering in complex environments. In addition, the customized architecture flexibly captures signal propagation under diverse LoS and NLoS conditions, allowing more precise modeling of RF interactions. 5.3. Case Study III: System Reconfiguration A reconfigurable wireless system enables dynamic optimization of wireless environments via signal re… view at source ↗

**Figure 9.** Figure 9: 3D classroom models for simulation purposes. We provide (a) the classroom dimensions along with (b) an image captured by the camera. 明德厚学求是创新华中科技大学电子信息与通信学院 (a) Rendered magnitude of the (c) Rendered roughness reflection coefficient (b) Rendered phase of the reflection coefficient [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗

**Figure 10.** Figure 10: Visualization of the rendered (a) magnitude and (b) phase of the reflection coefficient, as well as the rendered (c) roughness maps, obtained via inverse rendering of the classroom scene. H. Physics-Guided LoS/NLoS Signal Mixing. We decompose the LoS/NLoS signal synthesis into a direct LoS component and a scattered NLoS component rendered by RFIR. Specifically, the signal propagation process is modeled in… view at source ↗

read the original abstract

Neural rendering paradigms have recently emerged as powerful tools for radio frequency (RF). However, by entangling RF sources with scene geometry and material properties, existing approaches limit downstream manipulation of scene geometry, wireless system configuration, and RF reasoning. To address this, we propose a physically grounded RF inverse rendering (RFIR) framework that explicitly decouples RF emission, geometry, and material electromagnetic properties. Our key insight is an RF-aware bidirectional scattering distribution function, embedded into the Gaussian splatting paradigm as an RF rendering equation. Each Gaussian primitive is endowed with intrinsic physical attributes, including surface normals, material electromagnetic parameters, and roughness, and leveraged by a customized ray-tracing scheme to represent RF signal synthesis. The proposed RFIR generalizes three typical RF tasks: radar cross-section synthesis, received signal strength indicator prediction, and wireless scene editability. Experiments demonstrate significant performance advantages, underscoring the potential for wireless world modeling.

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

RFIR embeds an RF-aware BSDF in Gaussian splatting to decouple geometry and materials for RCS, RSSI, and scene editing tasks, but the ray-tracing fidelity for real EM effects is the open question.

read the letter

The main thing to know is that this paper takes Gaussian splatting and adds a custom RF-aware BSDF so each primitive carries normals, electromagnetic parameters, and roughness, then runs a tailored ray-tracer to synthesize signals. That decoupling is the explicit goal and it is applied to three tasks: radar cross-section synthesis, received signal strength prediction, and wireless scene editing. The abstract reports performance gains over prior entangled neural methods, which is the practical hook. What the work does cleanly is identify the manipulation limits of existing RF neural renderers and sketch a single framework that supports independent changes to geometry or materials. That is a reasonable step forward if the implementation delivers. The soft spot sits in the physics. Standard Gaussian splatting approximations tend to drop coherent effects like diffraction, interference, and phase accumulation that dominate at RF wavelengths, and the stress-test note flags exactly this gap. Without side-by-side results against full-wave solvers or anechoic measurements, the claimed advantages could reflect rendering shortcuts rather than accurate propagation. The abstract does not give equations or error bars, so it is hard to judge how much of the reported edge survives real validation. This paper is for the small group already working on neural methods for wireless environment modeling and sensing. A reader who wants editable RF scenes and is willing to test the physical claims themselves will get the most out of it. It is not yet ready for broad adoption, but the idea is distinct enough and the tasks are concrete enough that it deserves a serious referee to check the ray-tracing details and the experimental controls. I would send it to peer review rather than desk-reject.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes RFIR, a physically grounded inverse rendering framework for radio-frequency signals that decouples emission, geometry, and material electromagnetic properties. It embeds an RF-aware bidirectional scattering distribution function (BSDF) into the Gaussian splatting paradigm, endows each Gaussian primitive with surface normals, material EM parameters, and roughness, and employs a customized ray-tracing scheme to synthesize RF signals. The framework is claimed to generalize to three tasks—radar cross-section synthesis, received signal strength indicator prediction, and wireless scene editability—with experiments showing significant performance advantages.

Significance. If the RF rendering equation and Gaussian-based ray tracing prove physically faithful, the work could enable more interpretable and editable wireless world models, moving beyond entangled neural representations and supporting downstream manipulation of geometry, materials, and system configurations.

major comments (2)

[Abstract / RF rendering equation] The abstract asserts that the RF-aware BSDF and customized ray-tracing scheme yield accurate RF signal synthesis, yet the approach inherits standard Gaussian splatting approximations that omit coherent wave effects (phase, diffraction at wavelength-scale features, higher-order scattering). Without explicit comparison to full-wave solvers (e.g., FDTD or MoM) or anechoic measurements in the experiments, it is unclear whether reported performance gains reflect physical fidelity or rendering heuristics.
[Method (Gaussian attribute assignment) / Experiments] The central generalization claim (RCS synthesis, RSSI prediction, scene editability) rests on the assumption that assigning intrinsic physical attributes to Gaussian primitives produces parameter-free or physically consistent predictions. If material EM parameters and roughness are fitted per scene, the performance advantages may reduce to data-driven fitting rather than the claimed physical decoupling; this needs to be clarified with ablation on parameter count and cross-validation.

minor comments (2)

[Abstract] The term 'RF-aware bidirectional scattering distribution function' is introduced without a concise definition or reference to standard BSDF formulations in graphics or prior RF scattering models.
[Method] Notation for material electromagnetic parameters (permittivity, permeability, conductivity) and roughness should be standardized with a table or explicit equations to avoid ambiguity when readers compare to conventional EM literature.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our RFIR framework. We address each major comment below with clarifications on the physical approximations and the role of optimized attributes, while committing to revisions where appropriate.

read point-by-point responses

Referee: [Abstract / RF rendering equation] The abstract asserts that the RF-aware BSDF and customized ray-tracing scheme yield accurate RF signal synthesis, yet the approach inherits standard Gaussian splatting approximations that omit coherent wave effects (phase, diffraction at wavelength-scale features, higher-order scattering). Without explicit comparison to full-wave solvers (e.g., FDTD or MoM) or anechoic measurements in the experiments, it is unclear whether reported performance gains reflect physical fidelity or rendering heuristics.

Authors: We acknowledge that the RF rendering equation inherits ray-tracing approximations from Gaussian splatting and does not model coherent wave phenomena such as phase interference, diffraction at sub-wavelength scales, or higher-order scattering. These omissions are deliberate for computational efficiency in the targeted high-frequency regimes of our tasks. The RF-aware BSDF provides physical grounding by incorporating material electromagnetic parameters, but we agree that the abstract could better qualify the approximation level. No direct FDTD or anechoic comparisons are present because the work prioritizes inverse rendering scalability over full-wave validation; we will revise the abstract and discussion sections to explicitly state the validity range of the approximations and note this as a limitation. revision: partial
Referee: [Method (Gaussian attribute assignment) / Experiments] The central generalization claim (RCS synthesis, RSSI prediction, scene editability) rests on the assumption that assigning intrinsic physical attributes to Gaussian primitives produces parameter-free or physically consistent predictions. If material EM parameters and roughness are fitted per scene, the performance advantages may reduce to data-driven fitting rather than the claimed physical decoupling; this needs to be clarified with ablation on parameter count and cross-validation.

Authors: The material EM parameters and roughness are optimized per scene during inverse rendering to fit observed signals, but they are constrained to physically interpretable ranges and directly represent quantities such as permittivity and conductivity. This structure enables the claimed decoupling, as demonstrated by the scene editability results where attribute modifications yield consistent RF changes without retraining. The approach is not parameter-free, but the physical parameterization supports generalization across tasks. We agree that further evidence is warranted and will add an ablation on the number of optimized parameters together with cross-scene validation experiments in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: framework proposal with independent empirical validation

full rationale

The paper introduces RFIR as a new decoupling of emission/geometry/materials via an RF-aware BSDF inside Gaussian splatting and a custom ray-tracer. No equations or self-citations are shown that reduce any claimed prediction (RCS synthesis, RSSI, scene editing) back to the fitted parameters by construction. The three-task generalization and performance claims rest on experiments rather than definitional identity or load-bearing self-citation chains. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 1 invented entities

The approach introduces new entities and assumptions about physical attribute assignment to Gaussians, with several free parameters for materials that are central to the decoupling claim.

free parameters (2)

material electromagnetic parameters
Assigned to each Gaussian primitive; likely fitted or optimized during inverse rendering.
roughness
Intrinsic physical attribute per Gaussian, used in RF rendering.

axioms (2)

domain assumption RF signal synthesis can be represented via ray-tracing on Gaussian primitives with physical attributes
Core to the customized ray-tracing scheme in the RF rendering equation.
domain assumption Gaussian splatting paradigm can be extended to RF without losing physical grounding
Embedded as RF rendering equation.

invented entities (1)

RF-aware bidirectional scattering distribution function (BSDF) no independent evidence
purpose: To model the scattering of RF signals based on material properties
Introduced as the key insight for the framework; no external evidence provided in abstract.

pith-pipeline@v0.9.0 · 5469 in / 1547 out tokens · 59213 ms · 2026-05-10T17:11:55.500499+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

47 extracted references · 8 canonical work pages · 1 internal anchor

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...
[2]

Amballa, C., Basu, S., Wei, Y.-L., Yang, Z., Ergezer, M., and Choudhury, R. R. Can NeRFs see without cameras? arXiv preprint arXiv:2505.22441, 2025

work page arXiv 2025
[3]

Rendering resources

Bitterli, B. Rendering resources. https://benedikt-bitterli.me/resources/, 2016. Accessed: December 17, 2025

2016
[4]

Computational methods for electromagnetic inverse scattering

Chen, X. Computational methods for electromagnetic inverse scattering. John Wiley & Sons, 2018

2018
[5]

RFCanvas : Modeling RF channel by fusing visual priors and few-shot RF measurements

Chen, X., Feng, Z., Sun, K., Qian, K., and Zhang, X. RFCanvas : Modeling RF channel by fusing visual priors and few-shot RF measurements. In Proceedings of the 22nd ACM Conference on Embedded Networked Sensor Systems, pp.\ 464--477, 2024

2024
[6]

M., and Falciasecca, G

Degli-Esposti, V., Fuschini, F., Vitucci, E. M., and Falciasecca, G. Measurement and modelling of scattering from buildings. IEEE transactions on antennas and propagation, 55 0 (1): 0 143--153, 2007

2007
[7]

Physics assisted deep learning for indoor imaging using phaseless Wi-Fi measurements

Deshmukh, S., Dubey, A., Ma, D., Chen, Q., and Murch, R. Physics assisted deep learning for indoor imaging using phaseless Wi-Fi measurements. IEEE Transactions on Antennas and Propagation, 70 0 (10): 0 9716--9731, 2022

2022
[8]

Relightable 3D Gaussians: Realistic point cloud relighting with BRDF decomposition and ray tracing

Gao, J., Gu, C., Lin, Y., Li, Z., Zhu, H., Cao, X., Zhang, L., and Yao, Y. Relightable 3D Gaussians: Realistic point cloud relighting with BRDF decomposition and ray tracing. In European Conference on Computer Vision, pp.\ 73--89. Springer, 2024

2024
[9]

X., and Luo, J

Han, X., Zheng, T., Han, T. X., and Luo, J. RayLoc : Wireless indoor localization via fully differentiable ray-tracing. arXiv preprint arXiv:2501.17881, 2025

work page arXiv 2025
[10]

Differentiable and learnable wireless simulation with geometric transformers

Hehn, T., Peschl, M., Orekondy, T., Behboodi, A., and Brehmer, J. Differentiable and learnable wireless simulation with geometric transformers. arXiv preprint arXiv:2406.14995, 2024

work page arXiv 2024
[11]

A., Cammerer, S., Nimier-David, M., Binder, N., Marcus, G., and Keller, A

Hoydis, J., Aoudia, F. A., Cammerer, S., Nimier-David, M., Binder, N., Marcus, G., and Keller, A. Sionna RT : Differentiable ray tracing for radio propagation modeling. In 2023 IEEE Globecom Workshops (GC Wkshps), pp.\ 317--321, 2023

2023
[12]

Learning radio environments by differentiable ray tracing

Hoydis, J., A \" t Aoudia, F., Cammerer, S., Euchner, F., Nimier-David, M., Ten Brink, S., and Keller, A. Learning radio environments by differentiable ray tracing. IEEE Transactions on Machine Learning in Communications and Networking, 2024

2024
[13]

SC-GS: Sparse -controlled Gaussian splatting for editable dynamic scenes

Huang, Y.-H., Sun, Y.-T., Yang, Z., Lyu, X., Cao, Y.-P., and Qi, X. SC-GS: Sparse -controlled Gaussian splatting for editable dynamic scenes. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 4220--4230, 2024

2024
[14]

Learnable wireless digital twins: Reconstructing electromagnetic field with neural representations

Jiang, S., Qu, Q., Pan, X., Agrawal, A., Newcombe, R., and Alkhateeb, A. Learnable wireless digital twins: Reconstructing electromagnetic field with neural representations. IEEE Open Journal of the Communications Society, 2025

2025
[15]

Maximizing parallelism in the construction of BVHs , Octrees , and K-d trees

Karras, T. Maximizing parallelism in the construction of BVHs , Octrees , and K-d trees. In Proceedings of the Fourth ACM SIGGRAPH/Eurographics Conference on High-Performance Graphics, volume 6, pp.\ 33--37, 2012

2012
[16]

3D Gaussian splatting for real‑time radiance field rendering

Kerbl, B., Kopanas, G., Leimk \"u hler, T., and Drettakis, G. 3D Gaussian splatting for real‑time radiance field rendering. ACM Transactions on Graphics (TOG), 42 0 (139): 0 1--14, 2023

2023
[17]

GS-IR : 3D Gaussian splatting for inverse rendering

Liang, Z., Zhang, Q., Feng, Y., Shan, Y., and Jia, K. GS-IR : 3D Gaussian splatting for inverse rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 21644--21653, 2024

2024
[18]

Scene structure based neural radio-frequency radiance fields for channel knowledge map construction

Liu, K., Jiang, W., and Yuan, X. Scene structure based neural radio-frequency radiance fields for channel knowledge map construction. IEEE Wireless Communications Letters, 15: 0 171--175, 2026

2026
[19]

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

Lu, H., Vattheuer, C., Mirzasoleiman, B., and Abari, O. NeWRF : A deep learning framework for wireless radiation field reconstruction and channel prediction. arXiv preprint arXiv:2403.03241, 2024

work page arXiv 2024
[20]

FERMI : Flexible radio mapping with a hybrid propagation model and scalable autonomous data collection

Luo, Y., Wang, Y., Chen, H., Wu, C., Lyu, X., Zhou, J., Ma, J., Zhang, F., and Zhou, B. FERMI : Flexible radio mapping with a hybrid propagation model and scalable autonomous data collection. arXiv preprint arXiv:2504.14862, 2025

work page arXiv 2025
[21]

Modeling the phase correlation of effective diffuse scattering from surfaces for radio propagation prediction with antennas at refined separation

Miao, Y., Gueuning, Q., and Oestges, C. Modeling the phase correlation of effective diffuse scattering from surfaces for radio propagation prediction with antennas at refined separation. IEEE transactions on antennas and propagation, 66 0 (3): 0 1427--1435, 2018

2018
[22]

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

Orekondy, T., Kumar, P., Kadambi, S., Ye, H., Soriaga, J., and Behboodi, A. WiNeRT : Towards neural ray tracing for wireless channel modelling and differentiable simulations. In The Eleventh International Conference on Learning Representations, 2023

2023
[23]

O zdogan, \

\"O zdogan, \"O ., Bj \"o rnson, E., and Larsson, E. G. Intelligent reflecting surfaces: Physics , propagation, and pathloss modeling. IEEE Wireless Communications Letters, 9 0 (5): 0 581--585, 2019

2019
[24]

Interference-exact radiative transfer equation

Partanen, M., H \"a yrynen, T., and Oksanen, J. Interference-exact radiative transfer equation. Scientific Reports, 7 0 (1): 0 11534, 2017

2017
[25]

Physically based rendering: From theory to implementation

Pharr, M., Jakob, W., and Humphreys, G. Physically based rendering: From theory to implementation . MIT Press, 2023

2023
[26]

GIR : 3D Gaussian inverse rendering for relightable scene factorization

Shi, Y., Wu, Y., Wu, C., Liu, X., Zhao, C., Feng, H., Zhang, J., Zhou, B., Ding, E., and Wang, J. GIR : 3D Gaussian inverse rendering for relightable scene factorization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

2025
[27]

A generalized ray formulation for wave-optical light transport

Steinberg, S., Ramamoorthi, R., Bitterli, B., d'Eon, E., Yan, L.-Q., and Pharr, M. A generalized ray formulation for wave-optical light transport. ACM Transactions on Graphics (ToG), 43 0 (6): 0 1--15, 2024 a

2024
[28]

A free-space diffraction BSDF

Steinberg, S., Ramamoorthi, R., Bitterli, B., Mollazainali, A., D'Eon, E., and Pharr, M. A free-space diffraction BSDF . ACM Transactions on Graphics (TOG), 43 0 (4): 0 1--15, 2024 b

2024
[29]

M., Cenni, N., Fuschini, F., and Degli-Esposti, V

Vitucci, E. M., Cenni, N., Fuschini, F., and Degli-Esposti, V. A reciprocal heuristic model for diffuse scattering from walls and surfaces. IEEE Transactions on Antennas and Propagation, 71 0 (7): 0 6072--6083, 2023

2023
[30]

R., Li, H., and Torrance, K

Walter, B., Marschner, S. R., Li, H., and Torrance, K. E. Microfacet models for refraction through rough surfaces. In Proceedings of the 18th Eurographics Conference on Rendering Techniques, pp.\ 195--206, 2007

2007
[31]

C., and Ling, Z

Wang, F., Huang, Y., Feng, Z., Xiong, R., Li, Z., Wang, C., Mi, T., Qiu, R. C., and Ling, Z. Dreamer: Dual-RIS-aided imager in complementary modes. IEEE Transactions on Antennas and Propagation, 73: 0 4863--4878, 2025 a

2025
[32]

Wang, F., Mi, T., Wang, C., Xiong, R., Wang, Z., and Qiu, R. C. Source localization and power estimation through RISs: Performance analysis and prototype validations. IEEE Transactions on Wireless Communications, 25: 0 9406--9420, 2025 b

2025
[33]

Integrated sensing and communication channel modeling: A survey

Wei, Z., Jia, J., Niu, Y., Wang, L., Wu, H., Yang, H., and Feng, Z. Integrated sensing and communication channel modeling: A survey. IEEE Internet of Things Journal, 12 0 (12): 0 18850--18864, 2025

2025
[34]

Neural representation for wireless radiation field reconstruction: A 3D Gaussian splatting approach

Wen, C., Tong, J., Hu, Y., Lin, Z., and Zhang, J. Neural representation for wireless radiation field reconstruction: A 3D Gaussian splatting approach. arXiv preprint arXiv:2412.04832, 2024

work page arXiv 2024
[35]

WRF-GS : Wireless radiation field reconstruction with 3D Gaussian splatting

Wen, C., Tong, J., Hu, Y., Lin, Z., and Zhang, J. WRF-GS : Wireless radiation field reconstruction with 3D Gaussian splatting. In IEEE INFOCOM 2025-IEEE Conference on Computer Communications, pp.\ 1--10. IEEE, 2025

2025
[36]

and Jin, Y.-Q

Xu, F. and Jin, Y.-Q. Bidirectional analytic ray tracing for fast computation of composite scattering from electric-large target over a randomly rough surface. IEEE Transactions on Antennas and Propagation, 57 0 (5): 0 1495--1505, 2009

2009
[37]

C., and Ling, Z

Yang, H., Jin, Z., Wu, C., Xiong, R., Qiu, R. C., and Ling, Z. R-NeRF : Neural radiance fields for modeling RIS-enabled wireless environments. In GLOBECOM 2024-2024 IEEE Global Communications Conference, pp.\ 3859--3864. IEEE, 2024

2024
[38]

Generalizable Radio-Frequency Radiance Fields for Spatial Spectrum Synthesis

Yang, K., Chen, Y., and Du, W. GWRF : A generalizable wireless radiance field for wireless signal propagation modeling. arXiv preprint arXiv:2502.05708, 2025 a

work page internal anchor Pith review Pith/arXiv arXiv 2025
[39]

GSRF: Complex-valued 3D Gaussian splatting for efficient radio-frequency data synthesis

Yang, K., Dong, G., Du, W., Srivastava, M., et al. GSRF: Complex-valued 3D Gaussian splatting for efficient radio-frequency data synthesis. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025 b

2025
[40]

Scalable 3D Gaussian splatting-based RF signal spatial propagation modeling

Yang, K., Du, W., and Srivastava, M. Scalable 3D Gaussian splatting-based RF signal spatial propagation modeling. In Proceedings of the 23rd ACM Conference on Embedded Networked Sensor Systems, pp.\ 680--681, 2025 c

2025
[41]

Neilf: Neural incident light field for physically-based material estimation

Yao, Y., Zhang, J., Liu, J., Qu, Y., Fang, T., McKinnon, D., Tsin, Y., and Quan, L. Neilf: Neural incident light field for physically-based material estimation. In European conference on computer vision, pp.\ 700--716. Springer, 2022

2022
[42]

PhySG : Inverse rendering with spherical Gaussians for physics-based material editing and relighting

Zhang, K., Luan, F., Wang, Q., Bala, K., and Snavely, N. PhySG : Inverse rendering with spherical Gaussians for physics-based material editing and relighting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 5453--5462, 2021 a

2021
[43]

P., Deng, B., Debevec, P., Freeman, W

Zhang, X., Srinivasan, P. P., Deng, B., Debevec, P., Freeman, W. T., and Barron, J. T. Nerfactor: Neural factorization of shape and reflectance under an unknown illumination. ACM Transactions on Graphics (ToG), 40 0 (6): 0 1--18, 2021 b

2021
[44]

A unified RCS modeling of typical targets for 3GPP ISAC channel standardization and experimental analysis

Zhang, Y., Zhang, J., Gong, H., Hu, X., Zhang, J., Xing, H., Luo, S., Xiong, Y., Yu, L., Yuan, Z., Liu, G., and Jiang, T. A unified RCS modeling of typical targets for 3GPP ISAC channel standardization and experimental analysis. IEEE Journal on Selected Areas in Communications, pp.\ 1--1, 2025

2025
[45]

NeRF2 : Neural radio-frequency radiance fields

Zhao, X., An, Z., Pan, Q., and Yang, L. NeRF2 : Neural radio-frequency radiance fields. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking, pp.\ 1--15, 2023

2023
[46]

Physics-informed generalizable wireless channel modeling with segmentation and deep learning: Fundamentals , methodologies, and challenges

Zhu, E., Sun, H., and Ji, M. Physics-informed generalizable wireless channel modeling with segmentation and deep learning: Fundamentals , methodologies, and challenges. IEEE Wireless Communications, 2024 a

2024
[47]

GS-ROR: 3D Gaussian splatting for reflective object relighting via SDF priors

Zhu, Z.-L., Wang, B., and Yang, J. GS-ROR: 3D Gaussian splatting for reflective object relighting via SDF priors. arXiv preprint arXiv:2406.18544, 2024 b

work page arXiv 2024