Precision Localization System Using LiDAR–map Structural Mismatch-based Slip Detection and AMCL Pose Covariance Adjustment

This paper proposes a real-time localization-recovery method for autonomous mobile robots based on LiDAR-map structural mismatch detection and adaptive adjustment of the Adaptive Monte Carlo Localization (AMCL) pose covariance. The proposed method detects slip-induced localization inconsistencies from structural discrepancies between the map and real-time LiDAR observations while preserving the original AMCL motion and measurement models. When a localization inconsistency is detected, the pose covariance is expanded according to the average corner-angle variation, thereby increasing the particle search region. Subsequent LiDAR-based measurement updates and resampling enable the AMCL pose estimate to reconverge. Because the original AMCL localization framework remains unchanged, the proposed method can be integrated with the existing ROS Navigation Stack as an independent module. The method was evaluated under three representative scenarios: a low-friction floor, an unsecured rug, and an external impact. The effects of the proposed method were evaluated separately using the actual robot trajectories measured by a ceiling-camera system and the AMCL-estimated trajectories obtained from the amcl_pose topic. In all three scenarios, the final-pose errors relative to the navigation goal decreased after applying the proposed method, as evaluated separately using the actual and AMCL-estimated poses. These results indicate that the proposed method educes final-pose deviations from the Navigation Goal under slip-induced localization inconsistencies and improves the robustness of AMCL-based navigation without requiring additional onboard sensors beyond the wheel encoder and LiDAR already used by AMCL.