SA-Radar
收藏资源简介:
SA-Radar是一个基于波形参数嵌入的雷达模拟方法,旨在通过自定义雷达属性生成可控和高效的雷达数据。它结合了生成雷达模拟和基于物理的雷达模拟的优势,通过3D U-Net模型ICFAR-Net来捕捉不同雷达配置下信号的变化。该方法避免了需要详细的雷达硬件规格,并允许在多种传感器设置下高效地模拟范围-方位-多普勒(RAD)张量。通过在多个下游任务上的广泛评估,SA-Radar的模拟数据展现出真实性和有效性,无论单独使用还是与真实数据结合使用,都能一致地提高模型性能。该框架还支持在新的传感器视角和编辑场景中进行模拟,展示了其在自动驾驶应用中作为通用雷达数据引擎的潜力。
SA-Radar is a radar simulation method based on waveform parameter embedding, which aims to generate controllable and efficient radar data by customizing radar attributes. It combines the advantages of generative radar simulation and physics-based radar simulation, and uses the 3D U-Net model ICFAR-Net to capture signal variations under different radar configurations. This method eliminates the need for detailed radar hardware specifications, and allows efficient simulation of range-azimuth-doppler (RAD) tensors across diverse sensor settings. Through extensive evaluations on multiple downstream tasks, the simulated data generated by SA-Radar has demonstrated authenticity and effectiveness; whether used alone or combined with real-world data, it consistently improves model performance. This framework also supports simulation in novel sensor viewpoints and edited scenarios, showcasing its potential as a general-purpose radar data engine in autonomous driving applications.



