Aiming at the problems of degradation and poor robustness of simultaneous localization and mapping(SLAM)of LiDAR in large scenes,a SLAM algorithm integrating the observations of LiDAR,inertial measurement unit(IMU)and real-time kinematic(RTK)is proposed.In order to unify the spatial datum of multi-source information and make the system get accurate initial optimization value,doppler velocimetry is used to obtain the global attitude during initialization.In order to adapt to the complex scene continuous positioning,a factor graph optimization framework is designed to dynamically adjust the respective weights and optimization modes,which integrates inter-frame constraints of LiDAR odometry,IMU pre-integration factor and RTK global constraints.In various urban scenes such as RTK occlusion and point cloud degradation,high-precision positioning and mapping effects are achieved.Experimental results show that compared with the same sensor information fusing with LIO-SAM,the proposed algorithm improves the positioning accuracy(root mean square error)by 65.8％in the RTK occlusion scenes.In addition,the system has good robustness and the positioning accuracy reaches decimeter level in the long-distance vehicle experiments.
维普
万方数据
