ART-SLAM: Accurate Real-Time 6DoF LiDAR SLAM

Real-time six degrees-of-freedom pose estimation with ground vehicles represents a relevant and well-studied topic in robotics due to its many applications such as autonomous driving 3D mapping. Although some systems already exist, they are either not accurate or struggle real-time settings. In this letter, we propose fast, modular LiDAR SLAM system for both batch online estimation. We first apply downsampling outlier removal, filter out noise reduce the size of input point clouds. Filtered clouds then used tracking, possibly aided by pre-tracking module, floor detection, optimize estimated trajectory. Efficient multi-steps loop closure optimization, achieved through g2o graph, last steps proposed pipeline. compare performance our state-of-the-art cloud-based methods, LOAM, LeGO-LOAM, A-LOAM, LeGO-LOAM-BOR, LIO-SAM HDL, show that achieves equal better accuracy can easily handle even cases without loops. The comparison is done evaluating trajectory displacement using KITTI (urban driving) Chilean (underground mine) datasets.