Rasterizing Wireless Radiance Field via Deformable 2D Gaussian Splatting

TL;DR

SwiftWRF introduces a deformable 2D Gaussian splatting framework to model Wireless Radiance Fields (WRF) under single-sided transceiver mobility. By representing the spatial spectrum with compact 2D Gaussian primitives and learning a continuous, position-conditioned deformation field, it achieves real-time spectrum synthesis via CUDA rasterization, dramatically reducing computation and memory compared with NeRF-based and 3D Gaussian methods. The two-stage training with annealed smoothing and the GOP residual scheme enable strong generalization to unseen TX/RX positions and robust downstream tasks, including AoA and RSSI prediction. Across real-world and synthetic indoor scenes, SwiftWRF delivers up to 500x speedups and substantial fidelity gains, demonstrating the practicality of GS-based WRF modeling for URLLC-like wireless applications.

Abstract

Modeling the wireless radiance field (WRF) is fundamental to modern communication systems, enabling key tasks such as localization, sensing, and channel estimation. Traditional approaches, which rely on empirical formulas or physical simulations, often suffer from limited accuracy or require strong scene priors. Recent neural radiance field (NeRF-based) methods improve reconstruction fidelity through differentiable volumetric rendering, but their reliance on computationally expensive multilayer perceptron (MLP) queries hinders real-time deployment. To overcome these challenges, we introduce Gaussian splatting (GS) to the wireless domain, leveraging its efficiency in modeling optical radiance fields to enable compact and accurate WRF reconstruction. Specifically, we propose SwiftWRF, a deformable 2D Gaussian splatting framework that synthesizes WRF spectra at arbitrary positions under single-sided transceiver mobility. SwiftWRF employs CUDA-accelerated rasterization to render spectra at over 100000 fps and uses a lightweight MLP to model the deformation of 2D Gaussians, effectively capturing mobility-induced WRF variations. In addition to novel spectrum synthesis, the efficacy of SwiftWRF is further underscored in its applications in angle-of-arrival (AoA) and received signal strength indicator (RSSI) prediction. Experiments conducted on both real-world and synthetic indoor scenes demonstrate that SwiftWRF can reconstruct WRF spectra up to 500x faster than existing state-of-the-art methods, while significantly enhancing its signal quality. The project page is https://evan-sudo.github.io/swiftwrf/.

Figures (17)

• Figure 1: Comparison with NeRF-based and GS-based WRF modeling approaches. NeRF-based solutions rely on computationally intensive volumetric rendering, while 3DGS suffers from dimensional redundancy due to an invariant projection. In contrast, our method achieves both efficiency and compactness by adopting a streamlined 2D Gaussian formulation tailored to the wireless domain.
• Figure 2: A toy example illustrating the wireless radiance field from a given TX-RX deployment.
• Figure 3: Illustration of RX (origin) antenna array (a) and spatial spectrum (b).
• Figure 4: 2DGS formulation and spectrum rasterization.
• Figure 5: 2DGS deformation for spectrum rendering under single-sided transceiver mobility.