DRR-LIO: A Dynamic-Region-Removal-Based LiDAR Inertial Odometry in Dynamic Environments
Yankun Wang, Weiran Yao, Bing Zhang, Jinyu Fu, Jian Yang, Guanghui Sun
Abstract
This article aims to solve the problem of ghost trail effect left by dynamic objects and improve the accuracy of localization and mapping purity. Based on the tightly coupled LiDAR inertial odometry via smoothing and mapping (LIO-SAM), a real-time dynamic region removal method is proposed to challenge the real high dynamic environment. A vertical voxel height descriptor is presented to accurately discriminate dynamic and static points. Inertial measurement unit (IMU) preintegration is used for initial pose estimation to preferentially remove dynamic objects. A weighted optimization strategy is designed to improve the accuracy of pose estimation. The proposed algorithms are tested on the self-collected dataset and the public UrbanLoco dataset, and they achieve good real-time performance, mitigating the effect of dynamic objects in various scenes. The results verify that the LiDAR-inertial-based dynamic region removal odometry (DRR-LIO) can well remove dynamic objects and improve localization accuracy.