Abstract
Owing to the low cost, small size, and convenience for installation, 2D LiDAR has been widely used in mobile robots for simultaneous positioning and mapping (SLAM). However, traditional 2D LiDAR SLAM methods have low robustness and accuracy in LiDAR-degenerated environments. To improve the robustness of the SLAM method in such environments, an innovative SLAM method is developed, which mainly includes two parts, i.e., the front-end positioning and the back-end optimization. Specifically, in the front-end part, the AKF (adaptive Kalman filter) method is applied to estimate the pose of the mobile robot, zero bias of acceleration and gyroscope, lever arm length, and the mounting angle. The adaptive factor of the AKF can dynamically adjust the variance of the process and measurement noises based on the residual. In the back-end part, a particle filter (PF) is employed to optimize the pose estimation and build the map, where the pose domain constraint from the output of the front-end is introduced in the PF to avoid mismatch and enhance positioning accuracy. To verify the performance of the method, a series of experiments is carried out in four typical environments. The experimental results show that the positioning precision has been improved by about 61.3-97.9%, 35.7-99.0%, and 43.8-93.0% compared to the Karto SLAM, Hector SLAM, and Cartographer, respectively.