Adaptive-LIO: Enhancing Robustness and Precision through Environmental Adaptation in LiDAR Inertial Odometry

Chengwei Zhao^#,1,2    Kun hu^#,3    Jie Xu^*,1,5    Lijun Zhao^*,1    Baiwen Han¹    Kaidi Wu³ Maoshan Tian⁴ Shenghai Yuan<sup>5

1</sup>HIT    ²Qisheng Intelligent Techology    ³CUMT(XuZhou)    ⁴UESTC    ⁵ NTU

(#-co-first authors) (*-corresponding authors)

 

About

The emerging Internet of Things (IoT) applications, such as driverless cars, have a growing demand for high-precision positioning and navigation. Nowadays, LiDAR inertial odometry becomes increasingly prevalent in robotics and autonomous driving. However, many current SLAM systems lack sufficient adaptability to various scenarios. Challenges include decreased point cloud accuracy with longer frame intervals under the constant velocity assumption, coupling of erroneous IMU information when IMU saturation occurs, and decreased localization accuracy due to the use of fixed-resolution maps during indoor-outdoor scene transitions. To address these issues, we propose a loosely coupled adaptive LiDAR-Inertial-Odometry named Adaptive-LIO, which incorporates adaptive segmentation to enhance mapping accuracy, adapts motion modality through IMU saturation and fault detection, and adjusts map resolution adaptively using multi-resolution voxel maps based on the distance from the LiDAR center. Our proposed method has been tested in various challenging scenarios, demonstrating the effectiveness of the improvements we introduce.

📝 Updates

• [2024.12] - Adaptive-lio is accepted to JIOT 2024. 🚀

Data Sequence without strict time synchronization and calibration

Platform

Qisheng L1 mobile chassis.

notes: The experimental platform is the Qisheng L1 mobile chassis, which features dual-wheel differential steering. It is equipped with a Velodyne VLP-16 and an external IMU, using an Intel Core i5 as the computing platform. Please note that our IMU and LiDAR have not undergone hardware time synchronization, and the extrinsics betweenthe LiDAR and IMU have not been strictly calibrated.

Dataset

  

Dataset	Full Name	Duration (s)	Distance (km)	LiDAR Type
QiSheng	industrial	485	00	Velodyne VLP-16
QiSheng	industrial2	414	00	Velodyne VLP-16
QiSheng	park1	479	00	Velodyne VLP-16
QiSheng	park2	315	0.0	Velodyne VLP-16

End-to-end errors

Dataset	DLIO	LIO-SAM	Point-lio	Fast-lio2	IG-lio	Ours
industrial1	4.485	13.935	x	11.778	21.815	2.4824
industrial2	0.185	2.467	1.778	9.547	1.737	0.107
parking1	1.81	2.27	3.164	5.53	1.77	0.492

Comparison with others

The performance of all algorithms on the parking1 dataset.

    

Z-axis errors. 

Over-range detection

Contributors