Graph Learning and Deep Unfolding for Electromagnetic Field Modelling and Antenna Systems

Traditional computational electromagnetics (CEM) methods—such as MoM, FEM, or FDTD—offer high fidelity, but struggle to scale with complex antenna arrays and large propagation networks. Recent advances in artificial intelligence allow electromagnetic phenomena to be represented as graphs, where nodes encode field samples, array elements, or propagation points. Property-Testing-Guided GNNs extend this further by ensuring theoretical robustness and explainability, while deep unfolding frameworks bridge optimization theory and neural inference. These innovations promise to reduce computational cost, enhance adaptability to sparse or incomplete data, and provide interpretable mappings between measured and simulated fields—advancing both design and diagnostics in antenna engineering. The emerging convergence between graph learning and deep unfolding is reshaping electromagnetic field analysis and antenna system design, with GNNs capturing complex spatial dependencies in fields and propagation networks, and deep unfolding architectures offering physics-informed interpretability for inverse and reconstruction tasks. However, further work is needed to translate these methods into broadly applicable tools for practical antenna and propagation scenarios.This Research Topic aims to gather innovative contributions that leverage graph-structured representations, topological reasoning, and algorithm unrolling techniques for antenna design, array calibration, inverse scattering, radar imaging, and signal reconstruction. The combination of the expressive power of GNNs with the mathematical transparency of deep unfolding can enable new paradigms for efficient, data-driven, and physically consistent modelling across electromagnetics, antennas, and propagation domains. We are particularly interested in approaches that address computational scalability, adaptability to complex or sparse datasets, theoretical robustness, and interpretability. We invite studies that advance the state of the art and demonstrate how AI-driven frameworks can support both measured and simulated field data in antenna system analysis.This Research Topic focuses on the intersection of graph-based and deep unfolding methods with electromagnetic modelling, antennas, and signal propagation, emphasizing scalability and interpretability. We welcome articles addressing, but not limited to, the following themes:- Graph-based electromagnetic field and propagation modelling- Property-Testing-Guided GNNs and topological reasoning for antenna array and propagation analysis- Deep unfolding and LISTA-style architectures for electromagnetic inverse problems, radar, and imaging- Physics-consistent neural reconstruction from incomplete or non-uniform measurements- Hybrid graph-optimization networks for array calibration and beamforming- Explainability, sparsity, and uncertainty quantification in AI-based electromagnetic modellingStudies integrating experimental validation or open datasets are particularly welcome. Topic editors: Manuel J. C. S. Reis, Russo Paola Source: https://www.frontiersin.org/research-topics/77384/graph-learning-and-deep-unfolding-for-electromagnetic-field-modelling-and-antenna-systems