基于线面路标特征辅助的激光SLAM

LiDAR SLAM Based on Linear and Planar Landmark Feature Assistance

针对现有的激光SLAM系统未充分利用环境中的结构化特征,提出了一种基于直线/平面路标特征优化的激光SLAM系统,用于提升系统定位精度,增加语义地图的有效信息。通过传统几何的方式去分割激光点云局部地图中的圆柱形杆状物体和地面,将杆状物体的轴线作为直线路标特征,地面作为平面特征路标,进行数据关联和残差计算,形成路标特征因子,结合现有激光SLAM的激光里程计因子和回环因子进行因子图优化,得到实时鲁棒位姿,同时也能优化路标特征。采用公开数据集和真实实验数据对定位精度进行了评估,所提基于路标特征优化的激光SLAM系统均优于其他方法。

Aiming at the current existing LiDAR SLAM systems that do not make full use of the structured features in the environment,this paper proposes a LiDAR SLAM system based on the optimization of linear-planar landmark features,which is used to improve the localization accuracy of the system and increase the effective information of the semantic map.Firstly,the cylindrical pole-shaped object and the ground in the LiDAR point cloud local map are segmented by traditional geometry method.The axis of the pole-shaped object is used as a linear landmark feature,and the ground is used as a planar landmark feature.Data association and residual computation are carried out to form the factors.Factor graph optimization is carried out by combining the LiDAR odometry factor and the loop factor of the existing LiDAR SLAM.The real-time robust localization is output and also optimize the landmark features.In the experiments,this paper uses public datasets and real experimental data to evaluate the localization accuracy.The proposed LiDAR SLAM system based on the optimization of landmark features outperforms other methods.