BIM-based Adaptive Monte Carlo Localization (AMCL) method for indoor mobile robot

Abstract

This paper presents an enhanced BIM-based Adaptive Monte Carlo Localization (AMCL) algorithm designed to overcome two critical technical barriers facing indoor mobile robotics: (1) the difficulty in acquiring a mobile robot’s priori information, and (2) insufficient localization accuracy in indoor environment. The study analyzes and extracts geometric and semantic information of interior space structure through a Building Information Model (BIM), then building multi-layer raster maps for localization and navigation applications. The semantic layer map serves as a priori information to fuse a mobile robot’s LiDAR and vision sensor measurement information during the mobile robot operation. The proposed BIM-AMCL method improves obtaining prior information required for localization from BIM and optimizes the localization process. The experimental results demonstrate that the BIM-AMCL method maintains errors within 10 cm for both global and tracking localization processes, and within 5° for orientation error. Compared to the lidar-only AMCL approach, the proposed method achieves a 41.69% reduction in positional error and a 60.71% decrease in angular error, with a 55.56% improvement in global localization speed. Furthermore, by leveraging BIM-derived prior knowledge, the robot attains immediate localization and navigation capabilities without requiring environmental exploration.