Research on Mobile Robot Localization and Mapping Method for Underground Long-Narrow Tunnels

Abstract

The underground roadway scene has few features, which is a typical degraded scene. The commonly used GPS localization methods on the ground can not be directly applied, which brings great challenges to the localization and mapping of mobile robots. This paper proposes to build a laser-UWB fusion slam system to realize the localization and mapping of underground environment to solve the problems. The EKF based method was used to fuse IMU observation information to construct UWB pseudo GNSS localization system. The localization constraints provided by UWB localization system were added to the pose optimization constraints using the graph optimization-based framework and provide a reliable initial estimate for scan matching. Then using the original point cloud preprocessing method, i.e., de-distortion, radius filtering, and voxel grid filtering, combines the optimized pose for accurate pose estimation and mapping. Finally, the field experiment shows that the proposed method is closer to the actual trajectory and there is no cumulative error. The odometer error in the X direction has been effectively controlled. Compared with the general slam method, it has a higher ability to resist degradation.