C. Oestges | Laurent Jacques

Fast, differentiable, gpu-accelerated ray tracing for multiple diffraction and reflection paths

Mon, 20 Apr 2026 00:00:00 +0000

Abstract: Ray tracing is a widely used method to model radio wave propagation in complex environments. However, traditional ray tracing tools are often not differentiable, preventing their use in gradient-based optimization and machine learning pipelines. We present a fast, differentiable, GPU-accelerated ray tracing framework supporting multiple diffraction and reflection paths, enabling efficient optimization of radio network parameters. TODO: abstract exact à coller depuis arXiv:2510.16172.

Note: Best Paper Award EuCAP 2026.

Demonstrating DiffeRT: An open-source library for optimizing radio networks with differentiable ray tracing

Mon, 26 May 2025 00:00:00 +0000

Abstract: 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. 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.

Comparing differentiable and dynamic ray tracing: Introducing the multipath lifetime map

Sun, 30 Mar 2025 00:00:00 +0000

Abstract: 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.

Radio Propagation Modeling in an Urban Scenario using Generative Ray Path Sampling

Wed, 01 Jan 2025 00:00:00 +0000

Abstract: Radio propagation modeling remains essential for telecommunications research, where Ray Tracing provides a robust tool for simulating electromagnetic interactions in complex environments. In our previous work, we proposed a Machine Learning-assisted Ray Tracing pipeline that uses a generative approach to efficiently sample valid ray paths, significantly reducing computational complexity. In this paper, we extend this work by validating the proposed model through a comparative analysis of coverage maps. Specifically, we evaluate the accuracy of Machine Learning-generated ray paths by comparing the derived coverage maps with those produced by exhaustive Ray Tracing simulations. The results show that our approach could achieve significant computational savings while maintaining high fidelity to the ground truth, demonstrating its potential for scalable applications in wireless network design.

DiffeRT2d: A Differentiable Ray Tracing Python Framework for Radio Propagation

Thu, 05 Sep 2024 00:00:00 +0000

Abstract: Ray Tracing (RT) is arguably one of the most prevalent methodologies in the field of radio propagation modeling. However, access to RT software is often constrained by its closed-source nature, licensing costs, or the requirement of high-performance computing resources. While this is typically acceptable for large-scale applications, it can present significant limitations for researchers who require more flexibility in their approach, while working on more simple use cases. We present DiffeRT2d, a 2D Open Source differentiable ray tracer that addresses the aforementioned gaps. DiffeRT2d employs the power of JAX (Bradbury et al., 2024) to provide a simple, fast, and differentiable solution. Our library can be utilized to model complex objects, such as reconfigurable intelligent surfaces, or to solve optimization problems that require tracing the paths between one or more pairs of nodes. Moreover, DiffeRT2d adheres to numerous high-quality Open Source standards, including automated testing, documented code and library, and Python type-hinting.

Min-Path-Tracing: A Diffraction Aware Alternative to Image Method in Ray Tracing

Wed, 01 Mar 2023 00:00:00 +0000

Abstract: For more than twenty years, Ray Tracing methods have continued to improve on both accuracy and computational time aspects. However, most state-of-the-art image-based ray tracers still rely on a description of the environment that only contains planar surfaces. They are also limited by the number of diffractions they can simulate, owing to the complexity introduced when solving for paths. In this paper, we present Min-Path-Tracing (MPT), an alternative to the well-known image method that can handle diffractions seamlessly, while also leveraging the possibility to use different geometries for surfaces or edges, such as parabolic mirrors. MPT uses implicit representations of objects to write the path finding challenge as a minimization problem. We further show that multiple diffractions can be important in some situations, which MPT is capable to simulate without increasing neither the computational nor the implementation complexity.