For long-duration operations in GPS-denied environments, accurate and repeatable waypoint navigation is an essential capability. While simultaneous localization and mapping (SLAM) works well for single-session operations, repeated, multi-session operations require robots to navigate to the same spot(s) accurately and precisely each and every time. Localization and navigation errors can build up from one session to the next if they are not accounted for. Localization using a global reference map works well, but there are no publicly available packages for quickly building maps and navigating with them. We propose a new architecture using a combination of two publicly available packages with a newly released package to create a fully functional multi-session navigation system for ground vehicles. The system takes just a few hours from the beginning of the first manual scan to perform autonomous waypoint navigation.

翻译：在GPS拒止环境中进行长期作业时，精确且可重复的航点导航是关键能力。虽然同时定位与地图构建（SLAM）在单次作业中表现良好，但重复的多批次作业要求机器人每次都能精确且一致地导航至同一位置。若未对定位与导航误差进行补偿，这些误差会在各次作业之间累积。基于全局参考地图的定位方法效果良好，但目前尚无公开可用的软件包支持快速建图及基于地图的导航。我们提出了一种新架构，结合两个公开可用软件包与一个新发布的软件包，构建了一套功能完整的地面车辆多批次导航系统。该系统从首次手动扫描开始，仅需数小时即可实现自主航点导航。