Ray tracing has become a standard for accurate radio propagation modeling, but suffers from exponential computational complexity, as the number of candidate paths scales with the number of objects raised to the power of the interaction order. This bottleneck limits its use in large-scale or real-time applications, forcing traditional tools to rely on heuristics to reduce the number of path candidates at the cost of potentially reduced accuracy. To overcome this limitation, we propose a comprehensive machine-learning-assisted framework that replaces exhaustive path searching with intelligent sampling via Generative Flow Networks. Applying such generative models to this domain presents significant challenges, particularly sparse rewards due to the rarity of valid paths, which can lead to convergence failures and trivial solutions when evaluating high-order interactions in complex environments. To ensure robust learning and efficient exploration, our framework incorporates three key architectural components. First, we implement an experience replay buffer to capture and retain rare valid paths. Second, we adopt a uniform exploratory policy to improve generalization and prevent the model from overfitting to simple geometries. Third, we apply a physics-based action masking strategy that filters out physically impossible paths before the model even considers them. As demonstrated in our experimental validation, the proposed model achieves substantial speedups over exhaustive search – up to 10× faster on GPU and 1000× faster on CPU – while maintaining high coverage accuracy and successfully uncovering complex propagation paths. The complete source code, tests, and tutorial are available at https://github.com/jeertmans/sampling-paths.

We present a fast, differentiable, GPU-accelerated optimization method for ray path tracing in environments containing planar reflectors and straight diffraction edges. Based on Fermat's principle, our approach reformulates the path-finding problem as the minimization of total path length, enabling efficient parallel execution on modern GPU architectures. Unlike existing methods that require separate algorithms for reflections and diffractions, our unified formulation maintains consistent problem dimensions across all interaction sequences, making it particularly suitable for vectorized computation. Through implicit differentiation, we achieve efficient gradient computation without differentiating through solver iterations, significantly outperforming traditional automatic differentiation approaches. Numerical simulations demonstrate convergence rates comparable to specialized Newton methods while providing superior scalability for large-scale applications. The method integrates seamlessly with differentiable programming libraries such as JAX and DrJIT, enabling new possibilities in inverse design and optimization for wireless propagation modeling. The source code is openly available at https://github.com/jeertmans/fpt-jax.

Extremely Large-scale MIMO (XL-MIMO) systems operating in Near-Field (NF) introduce new degrees of freedom for accurate source localisation, but make dense arrays impractical. Sparse or distributed arrays can reduce hardware complexity while maintaining high resolution, yet sub-Nyquist spatial sampling introduces aliasing artefacts in the localisation ambiguity function. This paper presents a unified framework to jointly characterise resolution and aliasing in NF localisation and study the trade-off between the two. Leveraging the concept of local chirp spatial frequency, we derive analytical expressions linking array geometry and sampling density to the spatial bandwidth of the received field. We introduce two geometric tools–Critical Antenna Elements (CAEs) and the Non-Contributive Zone (NCZ)–to intuitively identify how individual antennas contribute to resolution and/or aliasing. Our analysis reveals that resolution and aliasing are not always strictly coupled, e.g., increasing the array aperture can improve resolution without necessarily aggravating aliasing. These results provide practical guidelines for designing NF arrays that optimally balance resolution and aliasing, supporting efficient XL-MIMO deployment.

Free-space optical communications (FSOC) between ground and satellites are limited by cloud cover and atmospheric turbulence. In order to establish optimal locations for future optical ground stations by quantifying atmospheric impacts, several measurement campaigns are ongoing. This work presents measurements of cloud cover and optical turbulence collected in Louvain-la-Neuve, Belgium, from August 2024 to July 2025 (1 year). The measurements are compared with estimations from the Weather Research and Forecasting (WRF) software, highlighting the interest of numerical weather prediction simulations to derive accurate statistics for assessing the viability of FSOC at a given location. The benefit of adaptive optics correction for optimizing single-mode fiber coupling in simulated GEO satellite downlinks in Louvain-la-Neuve is also demonstrated, showing that, in the studied scenario, link availability is primarily limited by cloud cover. (c) 2026 Optica Publishing Group. All rights, including for text and data mining (TDM), Artificial Intelligence (AI) training, and similar technologies, are reserved.

This paper proposes a sub-antenna array beamsteering-domain channel estimation framework for millimeter-wave (mmWave) systems with hybrid beamforming. Unlike conventional array-domain methods that require access to per-antenna baseband, the proposed approach leverages joint transmitter (Tx) and receiver (Rx) beamsteering observations to enable high-resolution, digital-beamforming-like parameter estimation under practical hardware constraints. A generalized signal model is developed that incorporates measured beam patterns, including sidelobe and phase effects, ensuring a practical representation of hybrid transceivers. Building on this model, we adapt the classical SAGE and MUSIC algorithms to operate in the beamsteering domain, allowing accurate estimation of multipath angle and delay parameters without relying on discretized angular grids. The proposed methods are evaluated through extensive Monte Carlo simulations across different SNR levels, array configurations, and beamsteering resolutions, and are further validated on a 28 GHz hybrid beamforming testbed. Results demonstrate that the beamsteering-based SAGE achieves superior resolution in dense multipath environments, while beamsteering-based MUSIC provides robustness under reduced beamsteering diversity. These findings establish the proposed framework as a practical and effective solution for accurate mmWave channel parameter estimation with hybrid beamforming architectures.

The increased carrier frequencies envisioned for future vehicle-to-vehicle or vehicle-to-infrastructure (commonly abbreviated as V2X) radar and communication networks call for adequate propagation models valid in near-field transmissions. In this paper, we propose a new radar propagation model, encompassing popular models from the literature, namely the radar equation and the Geometrical Optics (GO) approximation. First, the scattered electromagnetic fields and radar cross-sections are analytically computed by modeling the radar target as an arbitrary curved rectangular plate. Secondly, when particularized to a flat plate, our new model directly highlights not only the link between the radar equation and the GO approximation, but also the limitations of those models. We finally propose a new model based on our analytical derivations, which unifies both approaches.

Atmospheric turbulence influence on optical wave propagation, referred to as optical turbulence, has long been studied for astronomical applications and is now being addressed for free-space optical communication links between ground and satellites. While challenges overlap, models developed for astronomical applications are not fully transferable to optical communications. This paper provides a literature review of optical turbulence models, i.e., models giving vertical profiles of the refractive index structure parameter Cn2, highlighting differences between astronomical and optical communication sites. It presents different classifications of available Cn2 models, based on the atmospheric layer they target and their necessary input parameters. Boundary layer Cn2 models are also addressed, and recent machine learning approaches for Cn2 modelling are discussed. Additionally, commonly used metrics for comparing Cn2 profiles are introduced. Therefore, this work provides important insights into optical turbulence model selection, enabling accurate site characterization and informed optical terminal design.

This paper presents a chirp-based framework for characterising aliasing in a bistatic Near-Field (NF) imaging system equipped with multidimensional antenna arrays. Extending monostatic formulations, we derive closed-form expressions for the maximum spatial frequency, enabling the analytical derivations of the conditions for aliasing-free image reconstruction. The framework also provides a geometric interpretation of aliasing based on the antenna array geometry, target position, and antenna element spacing. Numerical results corroborate theoretical findings and show that the aliasing-free region enlarges with smaller antenna spacing, greater target range, lower array dimensionality, and smaller arrays. These results enable more effective design of bistatic NF imaging systems.

Existing studies analyzing electromagnetic field (EMFE) in wireless networks have primarily considered downlink communications. In the uplink, the EMFE caused by the user's smartphone is usually the only considered source of radiation, thereby ignoring contributions caused by other active neighboring devices. In addition, the network coverage and EMFE are typically analyzed independently for both the uplink and downlink, while a joint analysis would be necessary to fully understand the network performance and answer various questions related to optimal network deployment. This paper bridges these gaps by presenting an enhanced stochastic geometry framework that includes the above aspects. The proposed topology features base stations modeled via a homogeneous Poisson point process. The users active during a same time slot are distributed according to a mixture of a Matérn cluster process and a Gauss-Poisson process, featuring groups of users possibly carrying several equipments. In this paper, we derive the marginal and meta distributions of the downlink and uplink EMFE and we characterize the uplink to downlink EMFE ratio. Moreover, we derive joint probability metrics considering the uplink and downlink coverage and EMFE. These metrics are evaluated in four scenarios considering BS, cluster and/or intracluster densifications. Our numerical results highlight the existence of optimal node densities maximizing these joint probabilities.

This proposal presents a demonstration of DiffeRT, an open-source Differentiable Ray Tracing toolbox designed for Machine Learning and optimization applications in the context of radio propagation. Built on the JAX framework for differentiable programming and the Equinox library as a Machine Learning framework, DiffeRT extends its predecessor, DiffeRT2d, to support 3D scenes with enhanced performance and higher- level electromagnetic field computation features. The toolbox is fully scalable to GPUs and TPUs, enabling fast and efficient ray tracing for complex scenarios. Unlike existing alternatives like Sionna, DiffeRT focuses exclusively on ray-tracing functionalities, offering both low-level and high-level APIs for tailored solutions. This demonstration will showcase DiffeRT's unique capabilities, highlighting its suitability for a range of Machine Learning and optimization workflows. The tool is accessible via the Python Package Index (PyPI), and its source code can be obtained from its public repository: https://github.com/jeertmans/DiffeRT.

Radio propagation modeling is essential in telecommunication research, as radio channels result from complex interactions with environmental objects. Recently, Machine Learning has been attracting attention as a potential alternative to computationally demanding tools, like Ray Tracing, which can model these interactions in detail. However, existing Machine Learning approaches often attempt to learn directly specific channel characteristics, such as the coverage map, making them highly specific to the frequency and material properties and unable to fully capture the underlying propagation mechanisms. Hence, Ray Tracing, particularly the Point-to-Point variant, remains popular to accurately identify all possible paths between transmitter and receiver nodes. Still, path identification is computationally intensive because the number of paths to be tested grows exponentially while only a small fraction is valid. In this paper, we propose a Machine Learning-aided Ray Tracing approach to efficiently sample potential ray paths, significantly reducing the computational load while maintaining high accuracy. Our model dynamically learns to prioritize potentially valid paths among all possible paths and scales linearly with scene complexity. Unlike recent alternatives, our approach is invariant with translation, scaling, or rotation of the geometry, and avoids dependency on specific environment characteristics.

With the increasing presence of dynamic scenarios, such as Vehicle-to-Vehicle communications, radio propagation modeling tools must adapt to the rapidly changing nature of the radio channel. Recently, both Differentiable and Dynamic Ray Tracing frameworks have emerged to address these challenges. However, there is often confusion about how these approaches differ and which one should be used in specific contexts. In this paper, we provide an overview of these two techniques and a comparative analysis against two state-of-the-art tools: 3DSCAT from UniBo and Sionna from NVIDIA. To provide a more precise characterization of the scope of these methods, we introduce a novel simulation-based metric, the Multipath Lifetime Map, which enables the evaluation of spatial and temporal coherence in radio channels only based on the geometrical description of the environment. Finally, our metrics are evaluated on a classic urban street canyon scenario, yielding similar results to those obtained from measurement campaigns.

The accurate estimation of multipath components in hybrid beamforming millimeter-wave systems is challenging due to limited spatial channel information and multiple beamforming sidelobes. High-resolution antenna array algorithms like SAGE and MUSIC, which typically rely on baseband signals from individual an-tennas, are impractical for architectures with fewer RF chains. This paper proposes a signal model that leverages baseband signals from multiple beam-steering directions to overcome hardware limitations. By adapting the SAGE and MUSIC algorithms to exploit joint transmitter (Tx) and receiver (Rx) beam-steering direction baseband sig-nals, we achieve effective multi path component parame-ter estimation. Monte Carlo simulations at 28 GHz show that both algorithms demonstrate strong performance, with the majority of absolute angular errors falling below 10 degrees. While MUSIC achieves higher pre-cision under certain conditions, its computational time increases with finer beamsteering resolution. In contrast, SAGE offers a more computationally efficient and robust solution, with a slight tradeoff in precision.

Reconfigurable Intelligent Surfaces (RISs) represent a key emerging technology in wireless communication systems, allowing for precise control over previously uncontrolled propagation environments. This paper investigates the integration and optimization of RISs to enhance coverage in indoor environments using NVIDIA's Sionna Ray-Tracing (RT) tool. The RIS is modeled as an array of reflective tiles with adjustable phase profiles. We analyze two phase profile approaches (gradient-based and distance-based) for steering reflected signals toward blind zones identified in the transmitter-only coverage map. Based on the fixed positions of the transmitter and RIS, we propose a clustering-based algorithm to determine sub-optimal target positions, focusing on the least covered areas in the transmitter-only coverage map. Additionally, we introduce a ray-tracing-based search algorithm to provide insights into a sub-optimal RIS size. Simulation results demonstrate that the proposed methods effectively enhance signal coverage and minimize blind zones, confirming their viability for real-world implementations in RIS-assisted communications.

Reconfigurable Intelligent Surfaces (RISs) have emerged as a promising technology for enhancing wireless communication performance in complex indoor environments. However, the effectiveness of RIS deployment depends on several key factors, including position, size, and beam-steering target points. Existing research often optimizes RIS phase profiles assuming fixed positions and sizes, limiting the degrees of freedom for performance improvement. To address this issue, this paper proposes a novel ray-tracing based joint optimization framework that simultaneously determines the RIS position, size, and target points to enhance indoor coverage, particularly in blind spots. The method first identifies low-power cells in the environment by setting a minimum path gain threshold for satisfactory signal quality. It then uses the K-means algorithm to cluster these cells into groups and assigns the centroids as the RIS beam-steering target points. Next, feasible RIS positions are selected to maintain line-of-sight (LoS) connections with both the transmitter and target points. The RIS size is iteratively adjusted to balance performance gains and hardware costs, selecting a near-optimal size based on a predefined performance improvement threshold. The proposed approach leverages ray-tracing simulations to provide physically consistent optimization based on realistic environments. Extensive simulations in an indoor office environment demonstrate that the proposed framework significantly improves coverage and signal quality in blind spots by systematically optimizing RIS deployment parameters. These findings highlight the potential of ray-tracing based optimization for practical RIS-aided wireless communication.

Traditional radar and integrated sensing and communication (ISAC) systems often approximate targets as point sources, a simplification that fails to capture the essential scattering characteristics for many applications. This paper presents a novel electromagnetic (EM)-based framework to accurately model the near-field (NF) scattering response of extended targets, which is then applied to three canonical shapes : a flat rectangular plate, a sphere and a cylinder. Mathematical expressions for the received signal are provided in each case. Based on this model, the influence of bandwidth, carrier frequency and target distance on localisation accuracy is analysed, showing how higher bandwidths and carrier frequencies improve resolution. Additionally, the impact of target curvature on localisation performance is studied. Results indicate that detection performance is slightly enhanced when considering curved objects. A comparative analysis between the extended and point target models shows significant similarities when targets are small and curved. However, as the target size increases or becomes flatter, the point target model introduces estimation errors owing to model mismatch. The impact of this model mismatch as a function of system parameters is analysed, and the operational zones where the point abstraction remains valid and where it breaks down are identified. These findings provide theoretical support for experimental results based on point-target models in previous studies.

Radar targets are traditionally modelled as point target reflectors, even in the near-field region. Yet, for radar systems operating at high carrier frequencies and small distances, traditional radar propagation models do not accurately model the scatterer responses. In this paper, a novel electromagnetic-based model is thus developed for the multistatic radar detection of a rectangular plate reflector in the near-field region. This model is applied to an automotive scenario, in which a linear antenna array is spread out at the front of a vehicle, and performs a radar measurement of the distance to the back of the vehicle ahead. Based on the developed received signal model, the maximum likelihood estimator of the range is designed. By exploiting the near-field target model, this estimator is shown to provide a significant gain with respect to traditional range estimators. The impact of the system and scenario parameters, i.e. the carrier frequency, bandwidth and distance to the target, is furthermore evaluated. This analysis shows that the radar resolution in the near-field regime is improved at high carrier frequencies, while saturating to the traditional bandwidth-dependent resolution in the far-field region.