This paper presents a novel approach to improve global localization and mapping in indoor drone navigation by integrating 5G Time of Arrival (ToA) measurements into ORB-SLAM3, a Simultaneous Localization and Mapping (SLAM) system. By incorporating ToA data from 5G base stations, we align the SLAM's local reference frame with a global coordinate system, enabling accurate and consistent global localization. We extend ORB-SLAM3's optimization pipeline to integrate ToA measurements alongside bias estimation, transforming the inherently local estimation into a globally consistent one. This integration effectively resolves scale ambiguity in monocular SLAM systems and enhances robustness, particularly in challenging scenarios where standard SLAM may fail. Our method is evaluated using five real-world indoor datasets collected with RGB-D cameras and inertial measurement units (IMUs), augmented with simulated 5G ToA measurements at 28 GHz and 78 GHz frequencies using MATLAB and QuaDRiGa. We tested four SLAM configurations: RGB-D, RGB-D-Inertial, Monocular, and Monocular-Inertial. The results demonstrate that while local estimation accuracy remains comparable due to the high precision of RGB-D-based ORB-SLAM3 compared to ToA measurements, the inclusion of ToA measurements facilitates robust global positioning. In scenarios where standard mono-inertial ORB-SLAM3 loses tracking, our approach maintains accurate localization throughout the trajectory.

翻译：本文提出了一种新颖方法，通过将5G到达时间测量值集成至ORB-SLAM3（一种同步定位与建图系统）中，以提升室内无人机导航的全局定位与建图性能。通过融合来自5G基站的ToA数据，我们将SLAM的局部参考坐标系与全局坐标系对齐，从而实现精确且一致的全局定位。我们扩展了ORB-SLAM3的优化流程，以集成ToA测量值并同步进行偏差估计，从而将固有的局部估计转化为全局一致估计。该集成有效解决了单目SLAM系统中的尺度模糊性问题，并增强了系统的鲁棒性，尤其在标准SLAM可能失效的挑战性场景中表现突出。我们使用五个真实室内数据集（通过RGB-D相机和惯性测量单元采集）对本方法进行评估，并利用MATLAB和QuaDRiGa模拟了28 GHz和78 GHz频率下的5G ToA测量值进行数据增强。我们测试了四种SLAM配置：RGB-D、RGB-D-惯性、单目以及单目-惯性。结果表明，尽管由于基于RGB-D的ORB-SLAM3相比ToA测量具有更高精度，局部估计精度保持相当水平，但ToA测量值的引入显著提升了全局定位的鲁棒性。在标准单目-惯性ORB-SLAM3发生跟踪丢失的场景中，我们的方法能够在整个轨迹上保持精确的定位。