다중 송신기 환경에서 무선 신호 강도 지도 재구성을 위한 물리 기반 신경망

📝 Abstract

Accurately mapping the radio environment (e.g., identifying wireless signal strength at specific frequency bands and geographic locations) is crucial for efficient spectrum sharing, enabling Secondary Users (SUs) to access underutilized spectrum bands while protecting Primary Users (PUs). While existing models have made progress, they often degrade in performance when multiple transmitters coexist, due to the compounded effects of shadowing, interference from adjacent transmitters. To address this challenge, we extend our prior work on Physics-Informed Neural Networks (PINNs) for singletransmitter mapping to derive a new multi-transmitter Partial Differential Equation (PDE) formulation of the Received Signal Strength Indicator (RSSI). We then propose ReVeal-MT (Reconstructor and Visualizer of Spectrum Landscape for Multiple Transmitters), a novel PINN which integrates the multi-source PDE residual into a neural network loss function, enabling accurate spectrum landscape reconstruction from sparse RF sensor measurements. ReVeal-MT is validated using real-world measurements from the ARA wireless living lab across rural and suburban environments, and benchmarked against 3GPP and ITU-R channel models and a baseline PINN model for a single transmitter use-case. Results show that ReVeal-MT achieves substantial accuracy gains in multi-transmitter scenarios, e.g., achieving an RMSE of only 2.66 dB with as few as 45 samples over a 370-square-kilometer region, while maintaining low computational complexity. These findings demonstrate that ReVeal-MT significantly advances radio environment mapping under realistic multi-transmitter conditions, with strong potential for enabling fine-grained spectrum management and precise coexistence between PUs and SUs.

💡 Analysis

Accurately mapping the radio environment (e.g., identifying wireless signal strength at specific frequency bands and geographic locations) is crucial for efficient spectrum sharing, enabling Secondary Users (SUs) to access underutilized spectrum bands while protecting Primary Users (PUs). While existing models have made progress, they often degrade in performance when multiple transmitters coexist, due to the compounded effects of shadowing, interference from adjacent transmitters. To address this challenge, we extend our prior work on Physics-Informed Neural Networks (PINNs) for singletransmitter mapping to derive a new multi-transmitter Partial Differential Equation (PDE) formulation of the Received Signal Strength Indicator (RSSI). We then propose ReVeal-MT (Reconstructor and Visualizer of Spectrum Landscape for Multiple Transmitters), a novel PINN which integrates the multi-source PDE residual into a neural network loss function, enabling accurate spectrum landscape reconstruction from sparse RF sensor measurements. ReVeal-MT is validated using real-world measurements from the ARA wireless living lab across rural and suburban environments, and benchmarked against 3GPP and ITU-R channel models and a baseline PINN model for a single transmitter use-case. Results show that ReVeal-MT achieves substantial accuracy gains in multi-transmitter scenarios, e.g., achieving an RMSE of only 2.66 dB with as few as 45 samples over a 370-square-kilometer region, while maintaining low computational complexity. These findings demonstrate that ReVeal-MT significantly advances radio environment mapping under realistic multi-transmitter conditions, with strong potential for enabling fine-grained spectrum management and precise coexistence between PUs and SUs.

📄 Content

Existing spectrum sharing frameworks, such as those implemented in the TV White Space (TVWS) database and Citizens Broadband Radio Service (CBRS) Spectrum Access System (SAS), rely heavily on traditional statistical models. However, such models struggle to accurately capture the realworld spectrum occupancy and do not generalize well enough to capture shadowing and fading caused by different kinds of terrain and environmental conditions, leading to conservative approaches that over-protect the primary users (PUs) and cause discrepancies in channel availability for spectrum re-use [1]- [3]. In the meantime, deterministic models such as ray tracing require precise characterization of the complete propagation environment such as vegetation, trees, buildings, and material properties. Any errors in accurately defining these site-specific characteristics can degrade the models’ accuracy. In addition, such deterministic models are computationally expensive to be useful for at-scale, online spectrum management in dynamic radio environments. The existing stochastic and deterministic models also typically require the transmitter’s operational parameters, such as Effective Isotropic Radiated Power (EIRP), transmitter location, and antenna characteristics, which may not be available in real-world scenarios (e.g., where strong privacy or military secrecy are desired). The aforementioned drawbacks call for new models that are generically applicable to diverse environments and that are highly accurate in capturing the impact of transmitters and environmental factors (e.g., vegetation, trees, and buildings) on receiver signal strength while not requiring comprehensive, highly accurate information about the transmitters and environment.

To address the above challenge, data-driven modeling via Spectrum Cartography (SC) offers a promising solution avenue. In SC, ground-truth wireless signal measurements from sparsely distributed RF sensors are used to accurately generate the Radio Environment Map (REM) in the geographical area of interest [1], [4]- [7]. In particular, SC treats radio environment mapping as an ill-posed inverse problem where transmitter location and RF parameters are not available, and SC uses the spatial relationship between measurements to regenerate REMs [8]- [10]. The generated REMs have a wide range of applications in wireless communications, for instance, dynamically identifying white spaces for efficient spectrum sharing, optimizing power control for interference management, and facilitating seamless handover [6], [7].

However, a critical limitation of most existing SC techniques, including our previous work [11], is their focus on environments with a single dominant transmitter. In real-world wireless networks, multiple transmitters often operate noncoherently within the same frequency band and geographic area. For instance, in a rural community leveraging TV White Spaces (TVWS) for broadband access, a single area may contain multiple secondary user (SU) transmitters: fixed wireless access points providing internet to farms, sensors for agricultural IoT monitoring soil moisture and crop health, and equipment for precision irrigation systems [12]. The signals from these diverse SUs, all operating in the same underutilized band, create a complex and overlapping radio landscape. This superposition results in intricate interference patterns and cochannel coverage that cannot be captured by models designed for a single, isolated transmitter. Accurately characterizing this multi-transmitter environment is therefore essential for reliable spectrum sharing, preventing harmful interference between SUs themselves, and is a critical prerequisite for the practical, fine-grained management of spectrum resources in next-generation rural wireless networks.

Despite their promises, existing methods for generating REMs have significant limitations. For instance, techniques such as kriging and tensor decomposition assume a uniform spatial structure, failing to capture complex variations in signal strength often observed in real-world scenarios. In addition, these models typically require dense data, leading to high computational and sensing costs [8], [13]. Similarly, while deep learning approaches are powerful for matrix or tensor completion tasks, they often lack interpretability and require vast amounts of labeled data, which are impractical to collect in real-world settings.

To fill the gap in radio environment mapping, we propose ReVeal-MT, an extended Physics-Informed Neural Network (PINN) architecture for blind spectrum cartography in the presence of multiple transmitters. ReVeal-MT builds on our earlier single-transmitter work by deriving a new PDE formulation for received signal strength under multiple transmitters and incorporating this PDE as a physics-based constraint into the learning framework. This innovation enables ReVeal-MT to model the spatial superposition of signals while capturing real-world var

This content is AI-processed based on ArXiv data.