基于时间距离-熵减策略的同步定位与地图构建算法

Synchronous Localization and Mapping Algorithm Based on Time Distance-entropy Reduction Strategy

针对LIO-SAM(lidar inertial odometry-smoothing and mapping)算法在小范围室外建图时后端基于距离和时间阈值的回环检测存在误检测、重复检测的问题,本文提出一种基于时间距离—熵策略改进后端回环检测的TDE-LIO-SAM(LIO-SAM based on time distance-entropy reduction strategy)算法。该算法首先通过距离阈值和时间阈值初步筛选出备选的回环,然后根据回环点云的熵值减少量来确定最终的匹配回环,同时考虑到TDE-LIO-SAM在实际工程中的应用,引入了一种针对六轴IMU(惯性测量单元)快速重力对齐方法,使LIO-SAM在使用6轴IMU时也能快速进行重力对齐,进一步提高了算法对不同规格传感器的适用性。在开源KITTI数据集和自采数据集下进行实验,实验结果表明改进后TDE-LIO-SAM算法相较于LIO-SAM算法回环误匹配数量显著减少,定位精度提升8%以上,在KITTI07数据集上绝对定位精度达到5 cm以内。

In this study,we propose an improved back loop detection TDE-LIO-SAM(lidar inertial odometry-smoothing and mapping based on time distance-entropy reduction strategy)algorithm to resolve the issues of false detection and repeated back loop detection in LIO-SAM algorithm in small-scale outdoor mapping.The proposed algorithm is based on the distance and time threshold of the LIO-SAM algorithm.It preliminarily selects the optional loop through the distance threshold and time threshold and then determines the final matching loop according to the entropy reduction of the loop point cloud.With respect to the application of TDE-LIO-SAM in practical engineering,a fast gravity alignment method for a six-axis inertial measurement unit is introduced so that the algorithm can quickly align gravity.This further improves the applicability of the algorithm to different sensor specifications.Our findings of the experiments on the open-source KITTI dataset and self-collected dataset show that the improved TDE-LIO-SAM algorithm significantly reduces the number of loop mismatches compared with the LIO-SAM algorithm.In addition,the positioning accuracy of the improved TDE-LIO-SAM algorithm is improved by more than 8%,and its absolute positioning accuracy on the KITTI07 dataset is less than 5 cm.