Remove Dynamic Objects for Mapping in the Warehouse Environment

This paper proposes a mapping algorithm tailored for the warehouse environment to overcome the limitations of existing mapping algorithms in handling dynamic point clouds. In the warehousing environment, the presence of numerous moving objects, such as pedestrians and mobile robots, generates dynamic point clouds that adversely impact point cloud matching, thereby compromising the quality of the point cloud map. Taking advantage of the structural and highly repetitive nature of the warehouse environment, we use the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to classify point clouds obtained from laser radar scans. Furthermore, based on the variations in the positions of centroids for each point cloud class, this paper achieves real-time discrimination of dynamic point clouds. To enhance the accuracy and speed of SLAM localization, we introduce lidar odometry based on incremental Normal Distributions Transform(NDT). Ultimately, relying on the pose output from the lidar odometry and the lidar information after dynamic point cloud removal, we construct a more precise map. This approach effectively addresses the challenges posed by dynamic point clouds in mapping, providing a reliable solution for real-time and accurate environmental perception. The reliability of the algorithm is validated through experimental verification.