Physics-Inspired Distributed Radio Map Estimation

TL;DR

PI-DRME tackles privacy and communication challenges in radio map estimation by introducing a physics-informed, distributed learning framework that decouples a globally shared pathloss representation from client-specific shadowing features. The approach uses a two-autoencoder architecture, model-based interpolation, gradient-based regularization, and a weighted update constraint to balance shared and local representations under heterogeneous data with no landscaping information. Across experiments on the RadioMapSeer dataset, PI-DRME consistently outperforms both standalone RME and FL-based methods, especially when client data are highly heterogeneous. This work enables more accurate, privacy-preserving spectrum cartography in complex environments by embedding domain knowledge into distributed learning for wireless sensing.

Abstract

To gain panoramic awareness of spectrum coverage in complex wireless environments, data-driven learning approaches have recently been introduced for radio map estimation (RME). While existing deep learning based methods conduct RME given spectrum measurements gathered from dispersed sensors in the region of interest, they rely on centralized data at a fusion center, which however raises critical concerns on data privacy leakages and high communication overloads. Federated learning (FL) enhance data security and communication efficiency in RME by allowing multiple clients to collaborate in model training without directly sharing local data. However, the performance of the FL-based RME can be hindered by the problem of task heterogeneity across clients due to their unavailable or inaccurate landscaping information. To fill this gap, in this paper, we propose a physics-inspired distributed RME solution in the absence of landscaping information. The main idea is to develop a novel distributed RME framework empowered by leveraging the domain knowledge of radio propagation models, and by designing a new distributed learning approach that splits the entire RME model into two modules. A global autoencoder module is shared among clients to capture the common pathloss influence on radio propagation pattern, while a client-specific autoencoder module focuses on learning the individual features produced by local shadowing effects from the unique building distributions in local environment. Simulation results show that our proposed method outperforms the benchmarks in achieving higher performance.

Figures (3)

• Figure 1: PI-DRME framework: To handle the heterogeneity in RME, rather than averaging all local clients' models as in FL, a "globally shared encoder" is employed to capture a common representation of pathloss effects on radio propagation patterns. Meanwhile, a "client-specific autoencoder" is designed to learn the local shadowing effects within each client’s local environment. To enhance training efficiency, a weighted constraint regularization is applied as in contrastive learning wu2021contrastive, effectively drawing positive pairs closer together and pushing negative pairs away from the anchor.
• Figure 2: RMSE of different RME methods (a) under different cases and (b) with different numbers of clients in Case-1.
• Figure 3: Radio maps estimated by different methods (columns) given different cases of sensor measurements (rows).