JMSE, Vol. 12, Pages 2311: A Laser-Based SLAM Algorithm of the Unmanned Surface Vehicle for Accurate Localization and Mapping in an Inland Waterway Scenario

JMSE, Vol. 12, Pages 2311: A Laser-Based SLAM Algorithm of the Unmanned Surface Vehicle for Accurate Localization and Mapping in an Inland Waterway Scenario

Journal of Marine Science and Engineering doi: 10.3390/jmse12122311

Authors: Yang Wang Chao Liu Jiahe Liu Jinzhe Wang Jianbin Liu Kai Zheng Rencheng Zheng

It is important to improve the localization accuracy of the unmanned surface vehicle (USV) for ensuring safe navigation in an inland waterway scenario. However, the localization accuracy of the USV is affected by the limited availability of global navigation satellite system signals, the sparsity of feature points, and the high scene similarity in inland waterway scenarios. Therefore, this paper proposes a laser-based simultaneous localization and mapping (SLAM) algorithm for accurate localization and mapping in inland waterway scenarios. Inertial measurement unit (IMU) data are integrated with lidar data to address motion distortion caused by the frequent motion of the USV. Subsequently, a generalized iterative closest point (GICP) algorithm incorporating rejection sampling is integrated to enhance the accuracy of point cloud matching, involving a two-phase filtering process to select key feature points for matching. Additionally, a mixed global descriptor is constructed by combining point cloud intensity and distance information to improve the accuracy of loop closure detection. Experiments are conducted on the USV-Inland datasets to evaluate the performance of the proposed algorithm. The experimental results show that the proposed algorithm generates accurate mapping and significantly improves localization accuracy by 25.6%, 18.5%, and 23.6% compared to A-LOAM, LeGO-LOAM, and ISC-LOAM, respectively. These results demonstrate that the proposed algorithm achieves accurate localization and mapping in an inland waterway scenario.