Large-Scale Point Cloud Map Merging Algorithm via Planarity-Adaptive Hybrid Error Fusion

In high-risk environments such as recent disaster sites and underground structures, it is important to build accurate, large-scale, three-dimensional point cloud maps by integrating areas explored individually by multiple autonomous robots.However, existing point cloud registration algorithms are susceptible to rapid degradation in registration performance, especially in degenerative environments with scarce geometric features such as corridors and tunnels.To solve this problem, we propose a hybrid registration algorithm that utilizes voxel-wise covariance to maintain fast computational performance efficiency while adaptively adjusting the matching error model according to the planarity of the local surface, resulting in robust performance in degenerate environments.Experimental results using the KITTI dataset and the degenerative environment dataset show that the proposed method has significantly improved registration accuracy and stability in the degenerative environment compared to the conventional method.The results of this study are expected to contribute to improving the accuracy and reliability of large-scale point cloud map merging in practical robotic exploration environments such as tunnels and disaster environments.