LiDAR–Visual–Inertial Multi-UGV Collaborative SLAM Framework

The collaborative execution of tasks by multiple Unmanned Ground Vehicles (UGVs) has become a development trend in the field of unmanned systems. Existing collaborative Simultaneous Localization and Mapping (SLAM) frameworks mainly employ methods based on visual–inertial or LiDAR–inertial. However, the use of C-SLAM based on these three types of sensors is relatively less common. Therefore, these systems cannot achieve robust and accurate global localization performance in real-world environments. In order to address this issue, a LiDAR–visual–inertial multi-UGV collaborative SLAM framework is proposed in this paper. The whole system is divided into three parts. The first part constructs a front-end odometry by integrating the raw information from LiDAR, visual, and inertial sensors, which provides the accurate initial pose estimation and local mapping of each UGV for the collaborative system. The second part utilizes the similarity of different local mappings to form a global mapping of the environment. The third part achieves global localization and mapping optimization for multi-UGV localization system. In order to verify the effectiveness of the proposed framework, a series of real-world experiments have been conducted. Over an average trajectory length of 237 m, the framework achieves a mean Absolute Pose Error (APE) of 1.49 m and Relative Pose Error (RPE) of 1.68° after the global optimization. The experimental results demonstrate that the proposed framework achieves superior collaborative localization and mapping performance, with the mean APE reduced by 5.4% and mean RPE reduced by 1.4% compared to other methods.