A Novel Lidar Inertial Odometry with Moving Object Detection for Dynamic Scenes

Abstract

Simultaneous localization and mapping(SLAM) is an essential technology for robots to achieve motion planning and trajectory prediction in unknown environments. However, most SLAM algorithms do not consider moving objects, which makes it difficult to work robustly in highly dynamic scenes. We propose a dynamic lidar inertial odometry, building on FAST-LIO2, which can detect and remove moving objects. Our method detects and deletes dynamic objects by calculating the similarity scores of the corresponding clusters in multiple scans. If the position of an object changes between multiple scanning frames, the object is a dynamic object and should be removed. Therefore, it can enable the slam system to create a consistent map even in a highly dynamic environment. We designed and conducted extensive experiments and evaluations on the Urbanloco dataset. The experimental results show that our method can detection and remove moving objects in dynamic scenarios and improve the localization accuracy and robustness of the SLAM system.