A LiDAR SLAM system for dense forest mapping with iterated motion distortion correction

This paper presents a LiDAR SLAM system leveraging LiDAR-Inertial Odometry (LIO) that incorporates iterative motion distortion correction. The majority of contemporary LIO techniques include motion distortion correction capabilities that can rectify LiDAR point cloud distortions caused by rapid movements, facilitated by IMU preintegration. Unlike these approaches, our method uniquely conducts the distortion correction in an iterative manner. This is crucial as IMU biases, which can affect the preintegration phase, are updated throughout the optimization process. Our approach systematically performs scan matching between LiDAR and local map point clouds, alongside motion distortion correction, while concurrently updating the IMU biases. Experimental validation on the Newer College Dataset demonstrates that our method surpasses existing state-of-the-art techniques, primarily due to our iterative correction strategy. Furthermore, we present a SLAM framework based on our LIO method, incorporating a novel point removal technique for mapping dense forest environments. We have created our own dataset for this purpose, and qualitative assessments confirm the superior mapping accuracy of our system.