标题： GPS-DRIFT：使用IMU-GPS融合和不变滤波的海洋表面机器人定位 显示英文标题

标题： GPS-DRIFT: Marine Surface Robot Localization using IMU-GPS Fusion and Invariant Filtering

摘要： 本文提出了DRIFT不变状态估计框架的扩展，使得能够对GPS和IMU数据进行鲁棒融合，以准确估计姿态和航向。 最初为在海洋自主水面车辆（ASV）上测试和使用而开发，该方法也可用于其他移动系统。 在原始仅依赖本体感觉的DRIFT算法基础上，我们开发了一种保持对称性的传感器融合管道，利用不变扩展卡尔曼滤波器（InEKF）将全球定位更新直接集成到校正步骤中。 关键的是，我们引入了一种新的航向校正机制，结合GPS地速信息和IMU方向，克服了航向在推算导航中的固有不可观测性。 该系统部署并验证在定制的Blue Robotics BlueBoat上，但方法论的重点是融合外感受和本体感觉传感器的算法方法，以实现无漂移的定位和可靠的方位估计。 这项工作提供了在具有挑战性或GPS退化条件下的准确航向观测和定位的开源解决方案，并为未来的实验和比较研究奠定了基础。 显示英文摘要

摘要： This paper presents an extension of the DRIFT invariant state estimation framework, enabling robust fusion of GPS and IMU data for accurate pose and heading estimation. Originally developed for testing and usage on a marine autonomous surface vehicle (ASV), this approach can also be utilized on other mobile systems. Building upon the original proprioceptive only DRIFT algorithm, we develop a symmetry-preserving sensor fusion pipeline utilizing the invariant extended Kalman filter (InEKF) to integrate global position updates from GPS directly into the correction step. Crucially, we introduce a novel heading correction mechanism that leverages GPS course-over-ground information in conjunction with IMU orientation, overcoming the inherent unobservability of yaw in dead-reckoning. The system was deployed and validated on a customized Blue Robotics BlueBoat, but the methodological focus is on the algorithmic approach to fusing exteroceptive and proprioceptive sensors for drift-free localization and reliable orientation estimation. This work provides an open source solution for accurate yaw observation and localization in challenging or GPS-degraded conditions, and lays the groundwork for future experimental and comparative studies.