Range-Constrained Pose Graph-based SLAM: Applications to Deployable Ranging Beacons for Unknown Environment Exploration

Lidar-based Simultaneous Localization and Mapping (SLAM) is a critical part of robotic exploration in unknown underground environments. The state-of-the-art algorithms for lidar SLAM suffers from drift, requiring loop closures to correct the errors. However, real-time lidar loop closure methods can be limited by a small search range, large computational load and susceptibility to false matches. Dropped ranging beacons provide an additional source of measurements to reduce localization drift. Existing use of ranging beacons assume known and distributed beacon locations, which is not applicable to unknown environments, where beacons need to be deployed from a robot. We propose two methods to use information from dropped ranging beacons in graph-based SLAM and show, in simulation, the challenges in doing so. We then propose an approach to use dropped beacons for robust loop closure detection to seed lidar loop closure. This approach is proven with operation in underground field tests, and is shown to delivery equivalent accuracy to brute-force local lidarbased loop detection, but at less than 5% of the computational expense.