Simultaneous Localization And Mapping With Coupled Lidar And Imu In Rugged Terrain

With the rapid development of artificial intelligence technology,autonomous driving technology of unmanned vehicle is widely used in urban traffic,national defense security and other neighborhood.Compared with the structured road surface of urban road,the autonomous driving technology of unmanned vehicle in rugged environment in the field is more challenging.Simultaneous localization and mapping(SLAM)is an important part of unmanned vehicle environment perception,which is used to realize the self-positioning of unmanned vehicle without satellite positioning signal and to build high-precision environment map.However,it is easy to reduce the accuracy of SLAM algorithm positioning and Environmental Mapping on rough road.Therefore,this paper studies the laser radar SLAM and grid map construction under rough road.In order to increase the positioning accuracy of lidar odometer in rough road,a lidar odometer calculation method in rough environment is proposed in this paper.The algorithm uses voxel filter to downsampling the original point cloud,and uses radius filter to eliminate outliers;Then,according to the characteristics of rugged environment,the point cloud is divided into edge points and plane points based on the curvature of the point cloud;In order to improve the accuracy of feature matching and allocate the two kinds of point clouds efficiently,this paper proposes a feature extraction method based on sub region;Based on the obtained edge points and plane points,the features of point line and point surface are further constructed.The point clouds of the front and back frames are associated by point line matching and point surface matching,and the constraint equation is constructed.The L-M algorithm is used to solve the odometer information.In order to reduce the vibration interference of unmanned vehicle caused by rough road,this paper uses the IMU output as the correction basis to eliminate the point cloud motion distortion,and constructs the IMU pre-integration model.Based on the IMU preintegration,the tightly coupled algorithm of IMU and lidar is established,which effectively improves the normal positioning accuracy in rough environment.In order to improve the mapping accuracy in rugged environment,this paper also proposes a sliding window mapping algorithm based on key-frame,which uses G-N optimization method to match the key-frame point cloud with the previous n key frames,so as to smooth the environmental mapping error caused by rugged road,and further reduce the SLAM system error through loop detection.Based on the obtained 3D point cloud map,firstly,the ground points in the map are removed by using the ground plane fitting algorithm,and then according to the Octree principle,the 3D probability map is obtained by using the Octomap library expansion,and then the 2D grid map is obtained by projection.Based on the KITTI dataset and the scene of structured road and rough road in real environment,a series of experiments are carried out to verify the positioning and mapping accuracy of the tightly coupled SLAM algorithm,and the path planning test is carried out based on 2D grid map.Experiments show that SLAM Algorithm in this paper has higher plane positioning accuracy and smaller error in normal direction,and can achieve highprecision self-positioning and establish accurate environment map to support the path planning of unmanned vehicles.The research results of this paper can provide technical reference and solutions for environmental perception in rugged environment.