Enhanced SLAM based on 2D LiDAR and RGB-D camera fusion for mobile robots navigation

Real-time mapping and dynamic navigation for mobile robots present significant challenges, particularly due to the limitations of 2D LiDAR in environmental representation and the constraints of using a single RGB-D camera. This paper introduces a novel mapping method that enhances the traditional ORB-SLAM2 system by integrating 2D LiDAR and RGB-D camera data using Bayesian estimation. This approach enables the construction of dense maps, OctoMaps, and grid maps, improving the completeness and practicality of the mapping process. Additionally, Cartographer-SLAM is incorporated into the enhanced ORB-SLAM2 framework to further refine mapping capabilities. Comparative tests using the publicly available TUM dataset show that the proposed method reduces absolute pose error by 51.16%, with mapping trajectories closely aligning with ground truth values. The camera tracking trajectory improves by 16.2%. Experimental results demonstrate that the novel algorithm provides clearer environmental representations, increased accuracy, and higher mapping success rates.