Multirobot cooperative SLAM method based on adaptively weighted fusion of LiDAR and vision

To address the challenges of insufficient mapping integrity, compromised real-time performance, and reduced robustness in visual Simultaneous Localization and Mapping (SLAM) systems for robots operating under varying lighting conditions, this study proposes a multi-robot cooperative SLAM method based on adaptive weighted LiDAR and visual fusion. Building upon the ORB-SLAM3 framework, the proposed approach integrates depth information from LiDAR to enhance the 3D perception capability of visual feature points. By improving the Extended Kalman Filter (EKF)algorithm, the method achieves efficient fusion of high-precision LiDAR ranging data and temporal information from visual sensors, thereby improving the pose estimation accuracy of individual robots. On this basis, combined with depth map generation techniques, a local dense point cloud map is constructed for each robot. Through a key frame matching mechanism and the Iterative Closest Point (ICP) registration algorithm, a coordinate transformation relationship is established between robots with overlapping fields of view. These local maps are then unified into a global coordinate system to accomplish the integration and construction of multi-robot maps. To validate the effectiveness of the proposed method, comprehensive experiments were conducted in the Gazebo simulation environment. The results demonstrate that the method significantly enhances the stability and environmental adaptability of the mapping process while maintaining real-time performance