Efficient Transmission of Radiomaps via Physics-Enhanced Semantic Communications

Yueling Zhou ∗ , Achintha Wijesinghe ⋆ , Yue Wang † , Songyang Zhang ∗ , Zhipeng Cai † · 2025

Read Efficient Transmission of Radiomaps via Physics-Enhanced Semantic Communications by Yueling Zhou ∗ , Achintha Wijesinghe ⋆ , Yue Wang † , Songyang Zhang ∗ , Zhipeng Cai † (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-2501-10654.txt.

Introduction

Yueling Zhou and co-authors send radio coverage maps as meanings rather than bits in this 2025 paper. A radiomap shows where wireless signal reaches, guiding how networks share spectrum. Edge devices must ship those maps to a central server over shaky links, and full maps are heavy.

Their framework extracts the map's semantics: how radio waves travel and what the surroundings look like. A depth-style encoding captures the propagation physics. The server then rebuilds the map with a conditional image generator. A federated variant trains the models across users without pooling private data. Tests recovered coverage maps accurately at a fraction of the usual bandwidth.

Catalog record

CREATORYueling Zhou ∗ , Achintha Wijesinghe ⋆ , Yue Wang † , Songyang Zhang ∗ , Zhipeng Cai †
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-10654

Browse the full library · About the Tower of Babel

Full text

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

Efficient Transmission of Radiomaps via Physics-Enhanced Semantic Communications Thanks: This work was supported in part by the National Science Foundation under Grants No. 2413622, No. 2231209, No. 2315596, No. 2244219, No. 2146497, No. 2431452; and in part by a grant from BoRSF under contract LEQSF(2024-27)-RD-B-03. (Corresponding author: Songyang Zhang.) Yueling Zhou ∗ , Achintha Wijesinghe ⋆ , Yue Wang † , Songyang Zhang ∗ , Zhipeng Cai † Abstract Enriching information of spectrum coverage, radiomap plays an important role in many wireless communication applications, such as resource allocation and network optimization. To enable real-time, distributed spectrum management, particularly in the scenarios with unstable and dynamic environments, the efficient transmission of spectrum coverage information for radiomaps from edge devices to the central server emerges as a critical problem. In this work, we propose an innovative physics-enhanced semantic communication framework tailored for efficient radiomap transmission based on generative learning models. Specifically, instead of bit-wise message passing, we only transmit the key “semantics” in radiomaps characterized by the radio propagation behavior and surrounding environments, where semantic compression schemes are utilized to reduce the communication overhead. Incorporating the novel concepts of Radio Depth Maps, the radiomaps are reconstructed from the delivered semantic information backboned on the conditional generative adversarial networks. Our framework is further extended to facilitate its implementation in the scenarios of multi-user edge computing, by integrating with federated learning for collaborative model training while preserving the data privacy. Experimental results show that our approach achieves high accuracy in radio coverage information recovery at ultra-high bandwidth efficiency, which has great potentials in many wireless-generated data transmission applications. I Introduction Next-generation wireless technologies, such as vehicle-to-everything (V2X) and integrated sensing and communications (ISAC), have experienced substantial advancements, leading to increasing demand of high-precision radiomaps for panoramic awareness of the radio frequency (RF) spectrum coverage and radio environments [ 1 ] . Characterizing the multi-dimensional distribution of RF power spectral density (PSD) across various locations, frequencies, and time, radiomap plays a vital role in spectrum management tasks, such as unmanned aerial vehicle (UAV) path planning [ 2 ] and three-dimensional (3D) environment reconstruction [ 3 ] . Usually, dense radiomaps are reconstructed from sparse observations collected by mobile sensors and user devices [ 4 ] , making radiomap estimation (RME) a critical challenge in modern wireless communications. However, most existing radiomap-assisted applications focus on RME with a global view at single snapshot, while ignoring the fact that sparse observations are collected by edge devices or driving test on-the-fly, and are then transmitted to the central server for data processing [ 5 ] . For example, in the drone-assisted emergency communications, UAVs collect PSD observations for 3D environment reconstruction and temporary base station deployment, where timely radiomap transmission boosts the efficiency of spectrum management [ 6 ] . In this sense, efficient transmissions of radiomaps from edge devices allow RME and radiomap-assisted network optimization in dynamic and complex spectrum environments. Fig. 1: Three types of radiomap transmission methods: 1) Type 1 is the original data transmission, which transmit the landscape image together with the PSD value and coordinates of each sensing point. 2) Type 2 is data-oriented communications, which first reconstruct the radiomaps at the edge devices and utilize deep learning models for JSCC. 3) Type 3 is our proposed method via semantic communications, which only transmits the most critical semantic information for radiomap reconstruction at the receiver end. Fig. 2: Proposed semantic communications for radiomap transmission (inference phase): 1) At the transmitter side, raw data including the landscape images and sparse PSD samples are collected by edge devices, which are then encoded into key “semantics” characterized by building segmentation, radio propagation models and base station locations. With semantic compression and additional JSCC, these features are transmitted to the receiver side for radiomap reconstruction; 2) At the receiver side, the decoded semantic features, together with the physics-processed radio depth map are inputted into the pre-trained cGAN for radiomap reconstruction, which can be further utilized for downstream radiomap-assisted spectrum management, such as outage map prediction. Traditional spectrum information transmission relies on the data-oriented communications [ 7 ] , which focus on the bit-wise message delivery. As shown in Fig. 1 , one type of transmission methods is the direct transmission of original sparse observations collected by mobile devices, together with the landscape information such as satellite images. Then, the RME is conducted at the receiver end (cloud server) for downstream spectrum management. However, due to the ignorance of correlations among different sparse samples and the high resolutions of auxiliary landscape data, the raw data sharing might result in heavy communication overheads, particularly in the resource-limited sensor networks. On the other hand, recent development of learning techniques enables RME in the edge device, after which the learning-based joint source and channel coding (JSCC) is utilized for radiomap transmission [ 8 , 9 ] , e.g. type-2 shown in Fig. 1 . Despite some successes in improving communication efficiency, there are still redundant information in radiomap transmission. For example, the PSD observations in different locations usually share similar radio propagation behavior within a given sensing region of the edge devices. How to efficiently transmit radiomaps in real-time, distributed spectrum management remain an open question. Recent advancements in semantic communications, as a new era to transmit information, motivate us to seek for efficient radiomap transmission in the dynamic spectrum environments. Instead of focusing on the bit-wise data transmission and recovery as in conventional data-oriented communications, semantic communications aim to convey the most critical “semantic” information for the accomplishment of downstream tasks [ 10 ] . Existing semantic communication frameworks usually leverage a straightforward use of deep learning techniques, where the semantic features are extracted in the transmitter side and then delivered to the receiver end for data reconstruction. Typical examples include those based on auto-encoders [ 11 ] , transformers [ 12 ] and diffusion models [ 13 ] . However, most existing semantic communication frameworks are designed for handling the human-generated contents, while few of them deal with the wireless-generated data. Different from the human-generated contents, the semantics of wireless-generated data shall be related to both physics models (e.g., radio propagation in RME) and auxiliary data (e.g., images). To the best of our knowledge, it is still a lack of research results to efficiently embed the physical model information with semantics communications in the literature. To fill the gap, we propose a novel physics-enhanced semantic communications for efficient transmission of radiomaps, illustrated as type-3 in Fig. 1 . Specifically, we first extract the key semantic information and predict the radio propagation models from sparse observation samples to characterize the radio propagation behavior. Then, different semantic compression schemes are explored for bandwidth compression. With a pre-trained conditional generative adversarial nets (cGAN) [ 14 ] shared at both transmitter and receiver, the dense radiomap can be reconstructed at the receiver end. Moreover, for privacy preserving in multi-user semantic communications, federated learning is utilized for collaborative model training. Our contributions are summarized as follows: 1. We are the first to introduce the physics-enhanced semantic communications for radiomap transmission, which significantly reduce the communication overhead and increase the accuracy of RME. 2. Thanks to the effective feature exploration in learning-based embedding, we provide

Read the complete text