Uand-SLAM: a multiagent collaborative SLAM method based on multisource factor optimization

Multi-agent collaborative SLAM plays a crucial role in swarm robot navigation and environmental perception. However, in unstructured and poorly lit complex scenarios, traditional methods often suffer from insufficient data fusion and registration errors, leading to a decline in collaboration performance. In contrast, mapping systems could benefit from rich 3D geometric solutions. Inspired by this, this paper proposes a multi-agent collaborative SLAM method based on multifactor optimization. By integrating IMU, odometry, LiDAR, and visual information, a multi-factor joint optimization framework is established to enhance the robustness and accuracy of collaborative SLAM. The proposed method also introduces a keyframe selection mechanism based on information entropy and a dynamic weight distribution strategy, further improving collaboration efficiency. Experimental results demonstrate that the method outperforms existing approaches in localization and mapping performance in multi-agent complex environments.