Terahertz Spatial Wireless Channel Modeling with Radio Radiance Field

John Song 1 , Lihao Zhang 1 , Feng Ye 2 , and Haijian Sun 1 · 2025

Read Terahertz Spatial Wireless Channel Modeling with Radio Radiance Field by John Song 1 , Lihao Zhang 1 , Feng Ye 2 , and Haijian Sun 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-06277.txt.

Introduction

John Song, Lihao Zhang, Feng Ye, and Haijian Sun tested a new way in 2025 to map wireless paths at very high frequencies. Terahertz links will sit at the heart of sixth-generation networks, with vast bandwidth. But their signals fade fast, bend little, and bounce in sharp ways, so old channel models fit poorly. Dense test grids are too costly. The team's fix builds a radio radiance field from a visual scene plus a few radio reads.

They first built a fine mock scene for such signals. The field rebuilt from sparse reads caught the main paths well. It gave full spatial channel data with no dense sampling. The authors say the route could scale to cheap maps for future networks.

Catalog record

CREATORJohn Song 1 , Lihao Zhang 1 , Feng Ye 2 , and Haijian Sun 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-06277

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Terahertz Spatial Wireless Channel Modeling with Radio Radiance Field John Song 1 , Lihao Zhang 1 , Feng Ye 2 , and Haijian Sun 1 Abstract Terahertz (THz) communication is a key enabler for 6G systems, offering ultra-wide bandwidth and unprecedented data rates. However, THz signal propagation differs significantly from lower-frequency bands due to severe free space path loss, minimal diffraction and specular reflection, and prominent scattering, making conventional channel modeling and pilot-based estimation approaches inefficient. In this work, we investigate the feasibility of applying radio radiance field (RRF) framework to the THz band. This method reconstructs a continuous RRF using visual-based geometry and sparse THz RF measurements, enabling efficient spatial channel state information (Spatial-CSI) modeling without dense sampling. We first build a fine simulated THz scenario, then we reconstruct the RRF and evaluate the performance in terms of both reconstruction quality and effectiveness in THz communication, showing that the reconstructed RRF captures key propagation paths with sparse training samples. Our findings demonstrate that RRF modeling remains effective in the THz regime and provides a promising direction for scalable, low-cost spatial channel reconstruction in future 6G networks. I Introduction Wireless communication technologies have rapidly evolved in the past decades, driven by the ever-increasing demand for higher data rates, lower latency, and improved efficiency. As wireless data traffic continues to surge, next-generation systems are pushing toward Terabit-per-second (Tb/s) wireless links [ 1 , 2 ] . Terahertz (THz) communication, operating in the broad spectrum between 0.1 and 10 THz, has emerged as a promising solution to meet this demand. THz systems offer significantly wider bandwidths and reduced interference due to their high directivity and short transmission ranges [ 3 ] . However, the physical characteristics of THz waves introduce significant challenges to wireless communication systems. First, due to the extremely short wavelengths, THz signals experience severe free-space path loss compared to lower frequency bands under the traditional single-input single-output (SISO) configuration. Fortunately, this limitation can be compensated by densely packing a large number of tiny THz antenna elements (on the order of hundreds) within a small physical area. As a result, high antenna array gains at both the transmitter (Tx) and receiver (Rx) become feasible with beamforming, making such gigantic multiple-input multiple-output (MIMO) architectures a defining feature of THz communication systems. Second, the propagation behavior of THz waves differs significantly from lower frequency bands such as sub-6 GHz or frequency range 3 (FR3, 7-24 GHz). In the lower bands, diffraction and specular reflection are the dominant mechanisms that ensure coverage in non-line-of-sight (NLoS) regions. In contrast, these two kinds of interactions are rare at THz frequencies. Instead, diffuse reflections (scattering) become more pronounced and can be partially leveraged for the NLoS region coverage. Nonetheless, both the gigantic MIMO configuration and the reliance on scattering NLoS paths impose stringent requirements on the accuracy of spatial channel knowledge. Traditional pilot-based channel estimation methods—using full digital beamformer, sending pilots, decoding the full channel matrix, and computing beamforming weights—become increasingly impractical to scale at THz due to the gigantic size of the antenna arrays. Although sparsity-aware approaches and hybrid beamforming have been proposed to simplify this process, they are essentially capturing the few dominant propagation paths in the environment. And the goal of beamforming in such scenarios is to adaptively steer array pattern lobes towards these dominant paths to the target users and suppress the interference along the paths to other non-target users. Given the increasing pilot overhead, the cost and complexity of operating full-digital or hybrid beamformers, and the computational burden of real-time beamforming optimization, there is a growing interest in implementing the spatial beamforming configuration, which aims to identify strong propagation paths and perform beamforming accordingly. For this purpose, traditional spatial channel modeling approaches, such as cluster-model-based approaches, dateset-style methods, and ray tracing, face limitations in generalizability, spatial continuity, or computational efficiency [ 4 , 5 , 6 , 7 , 8 ] . In contrast, radio radiance field (RRF) methods have recently emerged as promising alternatives, as they capture the underlying physical principles of radio propagation rather than merely fitting observed outcomes [ 9 , 10 , 11 ] . For example, in our prior work, RF-3DGS [ 11 ] , we proposed a novel spatial wireless channel modeling framework that reconstructs the scene geometry from visual data and learns a continuous RRF using sparse wireless measurements. This enables accurate and efficient reconstruction of the Spatial-CSI, incorporating the path loss, delay, angle of arrival and departure (AoA and AoD) of strong paths, with fast training speed and real-time querying ability. RF-3DGS offers high spatial resolution, low memory requirements, and sub-second inference latency, making it highly suitable for real-time applications. In this work, we further extend the RF-3DGS framework to address the unique challenges associated with modeling RRF in the THz band. Given that scattering NLoS paths are significantly attenuated at THz frequencies, accurately reconstructing the path loss becomes critical for enabling effective spatial beamforming. The proposed RF-3DGS+ framework estimates the full propagation path information based on scene geometry and leverages this to calibrate the path loss of each multipath component (MPC). Our main contributions are summarized as follows: • We analyze the propagation characteristics of THz waves in complex environments and demonstrate how the RRF framework can effectively represent such spatial wireless channels. • We propose RF-3DGS+, which addresses the ambiguity present in prior RRF-based methods by explicitly reconstructing full path information, including propagation distance and directional gains. • To validate our approach, we develop a fine-grained THz simulation environment and conduct detailed experiments, demonstrating the superior accuracy and efficiency of RF-3DGS+ in modeling THz spatial channels. II Related Work To model how radio waves propagate in 3D space, several categories of spatial channel modeling have been proposed. Cluster-based models, such as those in [ 4 , 5 ] , follow the geometry-based spatial-consistent MPC tracking approach. These models approximate radio wave propagation as rays under far-field assumptions and characterize key reflectors and scatterers in the environment, but they lack the fine-grained spatial resolution required for advanced 6G applications. Ray tracing methods [ 6 ] also rely on geometric optics but use detailed environment reconstructions to simulate propagation. While these methods can yield accurate Spatial-CSI, their performance heavily depends on the quality of 3D meshes and surface electromagnetic (EM) property modeling. Furthermore, ray tracing is computationally intensive, limiting its real-time applicability. Dataset-based methods [ 7 , 8 ] store precomputed Spatial-CSI data at discrete 3D grid points, obtained from ray tracing or measurement campaigns. These methods enable fast querying and high fidelity, but they suffer from high data collection costs, limited generalizability across scenarios, significant memory requirements, and complex interpolation of spatial attributes (e.g., AoA, AoD). To understand Thz propagation characteristics, extensive measurement-based studies have been conducted [ 12 , 13 ] . These works provide detailed physical insights into THz scattering behaviors, including scattering patterns and coefficients of different materials. However, such studies mainly serve as theoretical and simulation purposes, and their direct application to practical modeling remains limited. Recently, neural radiance field (NeRF)-based methods have been introduced into the radio domain. NeRF is a popular approach for 3D reconstruction in computer vision, which describes the distribution of EM energy in

Read the complete text