RadaRays: Real-time Simulation of Rotating FMCW Radar for Mobile Robotics via Hardware-accelerated Ray Tracing

TL;DR

RadaRays addresses the need for realistic, real-time radar simulation for mobile robotics by introducing a hardware-accelerated ray-tracing framework for rotating FMCW radars. It models reflections, refractions, scattering, and multi-path effects, producing polar images via energy-conserving ray tracing and BRDF-based materials, and exposes a Gazebo plugin for seamless integration. The approach demonstrates closer alignment to real radar measurements than lidar-based simulators, improves image-based similarity metrics, and enables rapid software prototyping for radar-based SLAM and navigation, while maintaining practical run times on modern RTX hardware. The authors also provide a benchmarking procedure and discuss challenges in complex geometries and object detection, outlining a path toward higher fidelity (e.g., full path tracing) and broader deployment in mobile robotics workflows, with a clear evaluation framework for future radar simulators.

Abstract

RadaRays allows for the accurate modeling and simulation of rotating FMCW radar sensors in complex environments, including the simulation of reflection, refraction, and scattering of radar waves. Our software is able to handle large numbers of objects and materials in real-time, making it suitable for use in a variety of mobile robotics applications. We demonstrate the effectiveness of RadaRays through a series of experiments and show that it can more accurately reproduce the behavior of FMCW radar sensors in a variety of environments, compared to the ray casting-based lidar-like simulations that are commonly used in simulators for autonomous driving such as CARLA. Our experiments additionally serve as a valuable reference point for researchers to evaluate their own radar simulations. By using RadaRays, developers can significantly reduce the time and cost associated with prototyping and testing FMCW radar-based algorithms. We also provide a Gazebo plugin that makes our work accessible to the mobile robotics community.

Figures (6)

• Figure 1: Left: a Husky robot is scanning a corridor with a Navtech CIR radar producing the polar image on the top left. Right: RadarRays simulation in a hand-modelled virtual scene. By applying material properties for radar wave reflections to every object we can closely replicate the real measurements.
• Figure 2: Standard reflection model of RadaRays. A blue incidence ray $\vec{v}_0$ passes from air medium (white) to another material (gray) that lets the wave travel with 0.4 m/ns speed. The surface normal $\vec{n}_s$ is visualized in red. The directions of reflection $\vec{v}_1$, orange, and transmission $T$, green, are computed by Snell's law. The energy of transmission $E_2$ and reflection $E_1$ are computed by Fresnel equations. $E_1$ is the total area from the orange curve to the surface. The transmission energy $E_2$ is represented by the area between the green curve and the surface.
• Figure 3: Multi path reflections are important for classic radar as in (a) top. However, radar focusing mechanisms such as in (b) reduce the impact of these reflections as sketched in (a) bottom.
• Figure 4: Three stages of RadaRays. a) The result after the ray tracing in Sec \ref{['sec:radarays:a']}. b) We apply post-processing noise to the perfect signals to model measurement noise. c) Perlin noise to model additional received signals that do not relate to a transmitted signal. d) The real Navtech CIR radar image recorded from the same position.
• Figure 5: Trajectories estimated by CFEAR for MulRan's DCC and KAIST sequences (columns) using simulated radar data as input (rows). From top to bottom: Lidar-like simulations, RadaRays using BRDF of \ref{['eq:brdf']}, RadaRays using Cook-Torrance (CT) BRDF, actual Navtech data. Each trajectory is colored blue at the beginning and yellow at the end. All units are given in meters.