LiDAR-Based SLAM under Semantic Constraints in Dynamic Environments

Facing the realistic demands of application environment robots, simultaneous localisation and mapping (SLAM) has gradually moved from static environments to complex dynamic environments, while traditional SLAM methods usually result in pose estimation deviations caused by errors data association due interference elements environment. This problem is effectively solved present study proposing a approach based on light detection ranging (LiDAR) under semantic constraints environments. Four main modules are used for projection point cloud data, segmentation, element screening, map construction. A LiDAR segmentation network SANet spatial attention mechanism proposed, which significantly improves real-time performance accuracy segmentation. selection algorithm designed with prior knowledge reduce elements. The results experiments conducted public datasets SemanticKITTI, KITTI, SemanticPOSS show that robustness proposed improved.