SenseRay-3D: Generalizable and Physics-Informed Framework for End-to-End Indoor Propagation Modeling

TL;DR

SenseRay-3D tackles the challenge of scalable indoor propagation modeling by predicting 3D path-loss heatmaps directly from RGB-D scans using a sensing-driven voxel representation and a SwinUNETR predictor. The framework embeds occupancy, material-aware reflection/transmission cues, and transmitter–voxel geometry into a unified input, then reconstructs path-loss via a residual of the physical FSPL baseline: $\hat{L}(t,v)=L_{fspl}(t,v)+\hat{L}_{env}(t,v)$ with $L_{fspl}(t,v)=20\log_{10}(d(t,v))+20\log_{10}(f)-147.55$. A comprehensive synthetic dataset combining 3D-FRONT, BlenderProc, and Sionna RT provides ground-truth path-loss heatmaps for training and benchmarking. Experiments show a mean absolute error of $4.27$ dB on unseen environments and an inference latency of $217$ ms per sample, demonstrating strong generalization, physical consistency, and practical feasibility for sense-driven indoor network design.

Abstract

Modeling indoor radio propagation is crucial for wireless network planning and optimization. However, existing approaches often rely on labor-intensive manual modeling of geometry and material properties, resulting in limited scalability and efficiency. To overcome these challenges, this paper presents SenseRay-3D, a generalizable and physics-informed end-to-end framework that predicts three-dimensional (3D) path-loss heatmaps directly from RGB-D scans, thereby eliminating the need for explicit geometry reconstruction or material annotation. The proposed framework builds a sensing-driven voxelized scene representation that jointly encodes occupancy, electromagnetic material characteristics, and transmitter-receiver geometry, which is processed by a SwinUNETR-based neural network to infer environmental path-loss relative to free-space path-loss. A comprehensive synthetic indoor propagation dataset is further developed to validate the framework and to serve as a standardized benchmark for future research. Experimental results show that SenseRay-3D achieves a mean absolute error of 4.27 dB on unseen environments and supports real-time inference at 217 ms per sample, demonstrating its scalability, efficiency, and physical consistency. SenseRay-3D paves a new path for sense-driven, generalizable, and physics-consistent modeling of indoor propagation, marking a major leap beyond our pioneering EM DeepRay framework.

Figures (10)

• Figure 1: Schematic representation of SenseRay-3D
• Figure 2: Detailed architecture of the SwinUNETR-based predictor used in SenseRay-3D, showing voxelized input features, hierarchical encoding and decoding, and reconstruction of the path-loss heatmap.
• Figure 3: Overview of the proposed dataset generation process
• Figure 4: Virtual RGB-D scanning results: (a) RGB image, (b) depth map, and (c) semantic segmentation map.
• Figure 5: Illustration of the dataset representation used for supervised training. Each sample contains voxelized input features: (a) occupancy grid describing scene geometry, (b) reflection feature (dB) derived from ITU-based material parameters, (c) transmission feature (dB), and (d) transmitter–voxel distance field (m). In addition, (e) the FSPL (dB) serves as the baseline, while (f) the ground-truth path-loss heatmap (dB) obtained from Sionna.