Research On Outdoor LiDAR Positioning Algorithm Assisted By Inertial Navigation

As the basis of automatic driving,autonomous positioning technology needs to meet the basic requirements of high reliability.The positioning technology that relies on external information such as satellites is often affected by the environment such as buildings and tunnels,resulting in low positioning accuracy or even positioning failure.Light Detection and Ranging(LiDAR)has attracted much attention because of its wide field of view,no influence by light and high resolution.At the same time,Simultaneous Localization and Mapping(SLAM)technologies based on LiDAR and Inertial Navigation Systems(INS)are developing rapidly due to their complementary characteristics.However,the current SLAM algorithm based on the fusion of LiDAR and INS still has problems such as low accuracy of sensor external parameter calibration and poor quality of point cloud data.Both of them will bring interference to the accuracy of the positioning algorithm.Therefore,the research on how to improve the accuracy of external parameter calibration and the quality of point cloud is of great significance.The main contents of this paper are as follows:Firstly,the common coordinate system and the working principle of related sensors in the inertial navigation assisted LiDAR positioning system are introduced.This paper introduces the hardware device performance and software system structure of the positioning system from hardware and software,as well as the simulation data set used to verify the positioning algorithm performance.The LIO-SAM localization algorithm is introduced,and compared with LOAM algorithm.It is proved that algorithm can improve the localization performance of the carrier in complex urban environment by introducing inertial derivative data.Secondly,the principle and realization method of time calibration and space calibration of lidar and integrated navigation system are introduced.According to the principle of space calibration and the analysis of the experimental results,the problems of low degree of freedom and low precision are found in the calibration between sensors in vehicle-mounted conditions,and a joint calibration algorithm for vehicle-mounted sensors is proposed.The algorithm realized fast rough calibration of rotation parameters based on a large range of trajectory,and completed full parameter calibration with the improved point cloud optimization scheme on the premise of known rough rotation parameters.Compared with the traditional calibration algorithm,the improved calibration algorithm is proved to be able to solve the problem of low accuracy of the calibration algorithm under vehicle-mounted conditions.Finally,after the sensor calibration is completed,a noise elimination module is introduced into the framework of SLAM system to improve the positioning accuracy.By analyzing the causes of point cloud noise and the disadvantages of traditional filtering algorithm,a real-time hierarchical filtering algorithm based on the distribution characteristics of point cloud is proposed.The algorithm uses surface voxels to realize adaptive fast coarse filtering of point cloud and combines with dension-based noise clustering algorithm to perform local fine filtering.At the same time,in order to reduce the calculation amount of fine filtering and ensure the real-time performance of the algorithm,the ideas of principal component analysis and selective downsampling are introduced to process the coarse filtering results before fine filtering.Based on KITTI open data set and self-test data set,it is proved that the improved framework of SLAM localization algorithm can effectively improve the localization accuracy from both qualitative and quantitative aspects.