结合正态分布变换与线面ICP的弹性激光SLAM算法
Resilient LiDAR SLAM Algorithm Based on Normal Distributions Transform and Line-Plane ICP
王彬 1章浙涛 1何秀凤1
作者信息
- 1. 河海大学地球科学与工程学院,江苏 南京,211100
- 折叠
摘要
激光雷达(light detection and ranging,LiDAR)同时定位与建图(simultaneous localization and mapping,SLAM)中的位姿估计依赖于高精度和高可靠性的扫描匹配算法.针对实时LiDAR里程计与建图(LiDAR odometry and mapping in real-time,LOAM)框架中的点到线和点到面的迭代最近点算法(iterative closest point,ICP)在非结构化场景中退化的问题,提出了用于识别非结构化场景的环境特征值(environmental feature values,EFV),并根据EFV弹性地选择用正态分布变换(normal distributions transform,NDT)进行粗配准,实现了 一种基于扫描匹配的弹性实时激光SLAM算法NDT-LOAM.实验结果表明,EFV可以有效区分非结构化场景,并给出了 EFV阈值的调试方法.定位与建图实验分析表明,所提算法相比LOAM等经典的纯激光SLAM算法,在精度以及可靠性上均有较大提升,室外定位精度可从米级提升至分米级,在面对手持数据时也不会建图失败,能够得到全局一致性地图.因此此算法具有很好的环境适应性,丰富和发展了面向复杂环境的SLAM方法.
Abstract
Objectives:Pose estimation of light detection and ranging(LiDAR)simultaneous localization and mapping(SLAM)relies on scan matching algorithm with high accuracy and reliability.Methods:Based on the iterative closest point(ICP)algorithm in LiDAR odometry and mapping(LOAM),we propose a re-silient real-time LiDAR SLAM algorithm where the normal distributions transform(NDT)is flexibly se-lected according to a so-called environmental feature value(EFV)for identifying unstructured scenes.Re-sults:Experimental results show that the EFV can effectively distinguish unstructured scenes,and the de-bugging method of EFV threshold is given.The analysis of localization and mapping experiments show that compared with the classic LiDAR SLAM algorithms such as LOAM,the proposed algorithm has a great improvement in accuracy and reliability,of which the outdoor accuracy can be obtained from meter level to decimeter level.Moreover,the method can build a map and obtain a global consistent map when facing handheld data.Conclusions:Therefore,the proposed method has good environmental adaptability,thus enriching and developing the SLAM method for complex environments.
关键词
LiDAR/SLAM/扫描匹配/NDT/线面ICP/弹性系统Key words
LiDAR/SLAM/scan matching/NDT/line-plane ICP/resilient system引用本文复制引用
基金项目
国家自然科学基金(41830110)
国家自然科学基金(42374014)
江苏省自然科学基金(BK20200530)
出版年
2024
NETL
NSTL
维普
万方数据