Precise localization and mapping are critical for achieving autonomous navigation in self-driving vehicles. However, ego-motion estimation still faces significant challenges, particularly when GNSS failures occur or under extreme weather conditions (e.g., fog, rain, and snow). In recent years, scanning radar has emerged as an effective solution due to its strong penetration capabilities. Nevertheless, scanning radar data inherently contains high levels of noise, necessitating hundreds to thousands of iterations of optimization to estimate a reliable transformation from the noisy data. Such iterative solving is time-consuming, unstable, and prone to failure. To address these challenges, we propose an accurate and robust Radar-Inertial Odometry system, RINO, which employs a non-iterative solving approach. Our method decouples rotation and translation estimation and applies an adaptive voting scheme for 2D rotation estimation, enhancing efficiency while ensuring consistent solving time. Additionally, the approach implements a loosely coupled system between the scanning radar and an inertial measurement unit (IMU), leveraging Error-State Kalman Filtering (ESKF). Notably, we successfully estimated the uncertainty of the pose estimation from the scanning radar, incorporating this into the filter's Maximum A Posteriori estimation, a consideration that has been previously overlooked. Validation on publicly available datasets demonstrates that RINO outperforms state-of-the-art methods and baselines in both accuracy and robustness. Our code is available at https://github.com/yangsc4063/rino.

翻译：精确的定位与建图是实现自动驾驶车辆自主导航的关键。然而，自运动估计仍面临重大挑战，尤其是在全球导航卫星系统失效或极端天气条件下。近年来，扫描雷达因其强大的穿透能力已成为一种有效的解决方案。然而，扫描雷达数据本身包含高水平的噪声，需要数百至数千次优化迭代才能从噪声数据中估计出可靠的变换。这种迭代求解方法耗时、不稳定且容易失败。为应对这些挑战，我们提出了一种精确且鲁棒的雷达惯性里程计系统RINO，该系统采用非迭代求解方法。我们的方法将旋转和平移估计解耦，并采用自适应投票方案进行二维旋转估计，在确保求解时间一致性的同时提高了效率。此外，该方法实现了扫描雷达与惯性测量单元之间的松耦合系统，并利用了误差状态卡尔曼滤波。值得注意的是，我们成功地从扫描雷达估计了位姿估计的不确定性，并将其纳入滤波器的最大后验估计中，这一考量先前常被忽视。在公开数据集上的验证表明，RINO在准确性和鲁棒性上均优于现有最先进方法和基线。我们的代码可在 https://github.com/yangsc4063/rino 获取。