A Globally Optimized Multi-Robot Localization and Mapping Method with Lightweight Communication and Efficient Collaboration Strategy

Aimed at the challenges of multi-robot collaborative localization and mapping in large-scale scenarios, a globally optimized multi-robot localization and mapping method with lightweight communication and efficient collaboration strategy CSGO-SLAM is proposed. This method tightly couples LiDAR and inertial measurement data to compensate for the shortcomings of individual sensors. To mitigate the high communication burden among robots due to frequent point cloud data interaction, a distributed framework is proposed, employing reflectivity-based lightweight descriptors for information interaction. Addressing the low efficiency of loop closure search and the significant registration errors in point cloud alignment during collaboration, an efficient search strategy is introduced for rapid search, and a coarse-to-fine two-stage point cloud registration method is developed to enhance registration accuracy. Pairwise consistent measurement is utilized to eliminate false loop closures, ensuring the efficient and accurate establishment of reliable loops. To tackle the accumulation of errors in multi-body collaborative localization, a factor graph approach is used to incorporate odometry, inertial measurements, and two types of loop closure observations for collaborative pose optimization. To achieve the fusion of high-precision dense global point cloud maps, the method unifies global loop closures to construct a nonlinear optimization model, dynamically optimizing the relationships of map transformations. Validation through experiments with KITTI dataset and multiple unmanned vehicle platforms confirms the effectiveness of the proposed method. The results demonstrate that our approach significantly enhances multi-robot collaborative localization accuracy and stability in complex large-scale scenarios, with a reduction of 9.52% in localization root mean square error and 20.88% in standard deviation of localization error.