WiNeRF: Measurement Constrained Radiance Fields for Actionable Wireless Channel Modeling

Saif Ur Rahman; Rafid Umayer Murshed; Anton Dmitriev; Cagri Tanriover; Rahul C. Shah; Elahé Soltanaghai

cs.NIAug 24, 2026

WiNeRF: Measurement Constrained Radiance Fields for Actionable Wireless Channel Modeling

Authors: Saif Ur Rahman, Rafid Umayer Murshed, Anton Dmitriev, Cagri Tanriover, Rahul C. Shah, Elahé Soltanaghai

Organizations: University of Illinois Urbana-Champaign · Intel

Abstract

Wireless embedded systems increasingly rely on wireless channel information for decision making, yet practical platforms operate under severe constraints, including few antennas, narrow bandwidth, and sparse, noisy measurements. While neural field based approaches inspired by Neural Radiance Fields (NeRFs) have recently been explored for continuous wireless channel modeling, existing approaches depend on dense measurements or external priors such as known geometry, visual context, or angle-of-arrival (AoA) information, limiting their practicality in real-world deployments. We present WiNeRF, a neural field framework that learns a spatially continuous, complex-valued wireless channel representation directly from sparse channel state information (CSI) collected by commodity WiFi devices. WiNeRF embeds intrinsic system constraints, such as antenna geometry, limited spatial resolution, and phase uncertainty, as inductive biases through a 3D conical wave sampling model, a multi-resolution implicit scene representation, and a differentiable optimization framework for complex-valued channel learning. Across diverse indoor environments with non-line-of-sight (NLoS) regions, WiNeRF achieves a median prediction SNR of 5.3 dB, outperforming prior neural baselines by 4.9 dB on average (approximately 3x higher prediction SNR), and produces a task-agnostic channel representation that can be directly reused in standard signal-processing pipelines, including beamforming, AoA estimation, and RSSI coverage mapping, without modifying existing hardware or wireless protocols.

Figures & tables

Figure 1. (a) WiNeRF combines cone-based sampling regions with a multi-resolution implicit scene representation of coarse room-scale grids (yellow) and fine local-detail grids (blue) to learn a continuous wireless field from sparse CSI measurements. (b) The learned field supports downstream tasks such as RSSI coverage mapping and AoA profiling.

Figure 2. (a) Prior NeRF-based wireless models sample along infinitesimally thin line rays, implicitly assuming arbitrarily fine angular resolution. In contrast, (b) WiNeRF replaces line rays with cone-based support regions that reflect finite antenna-array observability, enabling hardware-aware spatial aggregation.

Method	Wave Representation	Priors Need	Experimental Data	Channel Output
NeRF 2	Voxelized Rays	Dense Training Data	✓	Task-dependent (e.g. RSSI, spatial spectrum)
WiNeRT	Ray segments	Path parameters (AoA, ToF)	x	CIR
RF-Canvas	Tensorial Field	3D Geometry	✓	RSSI
NeWRF	Voxelized Ray	Multipath AoA	x	Complex CSI
WiNeRF	Hash-based Cones	–	✓	Complex CSI

Table 1. Evolution of Neural Field Methods for Wireless Applications

Figure 3. Given sparse spatial CSI measurements, WiNeRF reconstructs a continuous complex channel field through three modules: (1) Cone Sampling Wave Model , which defines cone-based support regions and sampled conical frustums around the receiver; (2) Implicit Scene Representation , which encodes the (x,y,z) samples using a multi-resolution hash grid; and (3) CSI Reconstruction Module , which predicts and aggregates complex CSI .

Figure 4. Array-dependent angular support in WiNeRF. The figure schematically compares two ULA configurations with different numbers of antennas: (a) higher-element ULA with finer directional observability and (b) lower-element ULA with coarser directional observability. WiNeRF adjusts cone-based support width accordingly. With finer observability, three example AoA s occupy distinct conical support regions; with coarser observability, the broader support regions are grouped into two nearby AoA s.

Figure 5. Angular support region for a 4-antenna ULA .

Figure 6. The full pipeline of CSI estimation via per-frustum prediction and aggregation.

Panel of Figure 7 — Figure 7. (a) Mobile platform used for CSI data collection with a 4-antenna WiFi router, Intel Realsense depth camera, and laptop mounted on a Roomba. (b-d) WiNeRF experimental environments.

Panel of Figure 8 — Figure 8. WiNeRF outperforms all baselines in channel estimation accuracy at unseen WiFi receiver locations, while accurately preserving both magnitude and phase characteristics.

Panel of Figure 9 — Figure 9. WiNeRF preserves phase coherence for practical MIMO tasks: (a) Beamforming and (b) AoA estimation, using predicted CSI with classical signal processing methods.

Figure 10. Multipath resolution results using estimated CSI; WiNeRF preserves accurate multipath AoA characterization

Figure 11. RSSI spatial coverage prediction in Conference Room. WiNeRF accurately reproduces room-scale coverage patterns with fine spatial detail matching ground truth, capturing both the LoS hotspot and NLoS attenuation at edges.

Panel of Figure 12 — Figure 12. WiNeRF Sensitivity Analysis

Explore similar work

A Geometric Algebra-informed NeRF Framework for Generalizable Wireless Channel Prediction
Apr 13, 2026Jingzhou Shen, Luis Lago Enamorado, Shiwen Mao +1Wireless Communications Radio Propagation Modeling
Knight Foundation School of Computing and Information Sciences, Florida International University, Miami, FL 33199, US · Department of Electrical and Computer Engineering, Auburn University, Auburn, AL 36849, US
GeNeRT: A Physics-Informed Approach to Intelligent Wireless Channel Modeling via Generalizable Neural Ray Tracing
Jun 23, 2025Kejia Bian, Meixia Tao, Shu Sun +2Neural Network Generalization Wireless Communications
School of Information Science and Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China · Shanghai Innovation Institute, Shanghai 200231, China
RSS map-assisted MIMO channel estimation in the upper mid-band under pilot constraints
Feb 18, 2026Alireza Javid, Nuria González-PrelcicChannel Estimation Wireless Communications
Department of Electrical and Computer Engineering, University of California San Diego, USA

Apr 13, 2026J/K move · Enter open · S save

A Geometric Algebra-informed NeRF Framework for Generalizable Wireless Channel Prediction

Jingzhou Shen, Luis Lago Enamorado, Shiwen Mao, Xuyu Wang

Knight Foundation School of Computing and Information Sciences, Florida International University, Miami, FL 33199, US · Department of Electrical and Computer Engineering, Auburn University, Auburn, AL 36849, US

Wireless Communications Radio Propagation Modeling Neural Radiance Fields

Open paper AskPDF arXiv

In this paper, we propose the geometric algebra-informed neural radiance fields (GAI-NeRF), a novel framework for wireless channel prediction that leverages geometric algebra attention mechanisms to capture ray-object interactions in complex propagation environments. Our approach incorporates global token representations, drawing inspiration from transformer architectures in language and vision domains, to aggregate learned spatial-electromagnetic features and enhance scene understanding. We identify limitations in conventional static ray tracing modules that hinder model generalization and address this challenge through a new ray tracing architecture. This design enables effective generalization across diverse wireless scenarios while maintaining computational efficiency. Experimental results demonstrate that GAI-NeRF achieves superior performance in channel prediction tasks by combining geometric algebra principles with neural scene representations, offering a promising direction for next-generation wireless communication systems. Moreover, GAI-NeRF greatly outperforms existing methods across multiple wireless scenarios. To ensure comprehensive assessment, we further evaluate our approach against multiple benchmarks using newly collected real-world indoor datasets tailored for single-scene downstream tasks and generalization testing, confirming its robust performance in unseen environments and establishing its high efficacy for wireless channel prediction.

Abstract

Explore similar work

A Geometric Algebra-informed NeRF Framework for Generalizable Wireless Channel Prediction

GeNeRT: A Physics-Informed Approach to Intelligent Wireless Channel Modeling via Generalizable Neural Ray Tracing

RSS map-assisted MIMO channel estimation in the upper mid-band under pilot constraints