Research On Low-speed Driverless Vehicle Localization And Mapping Based On Sensor Fusion

With the rapid development of autonomous driving technology,a real-time and robust autonomous positioning system based on Simultaneous Localization and Mapping(SLAM)technology is crucial for intelligent vehicles.However,SLAM systems based on a single sensor are difficult to operate robustly in complex environments.Therefore,SLAM technology based on sensor fusion has become the current development trend.This paper realizes real-time positioning and map construction of unmanned vehicles in complex environments by fusing low-cost solid-state lidar and inertial measurement unit(IMU)sensor information.Firstly,accurate laser point cloud data is obtained through solidstate lidar data preprocessing algorithms.Secondly,the feature points in the preprocessing point cloud are extracted by the curvature and principal component analysis method to establish the constraint relationship between the adjacent frame point clouds,and the transformed poses of the two frame point clouds are estimated by scan matching and the local feature map is constructed.Thirdly,the solid-state lidar and IMU measurements are fused in a tightly coupled manner through a sliding-window optimization algorithm.Finally,an accurate,real-time and robust unmanned vehicle localization information is obtained.The main research contents of this paper are as follows:Firstly,in order to reduce the amount of original point cloud data of solid-state lidar,the voxel filtering method is used to downsample the original point cloud,while retaining the feature information in the point cloud.In order to reduce the influence of dynamic target point cloud on point cloud matching,the method of ground segmentation and clustering is used to extract the small target point cloud information in the point cloud.In the point cloud matching process,the information from the small target point cloud is not used to reduce the influence of dynamic targets on the point cloud matching.Secondly,in order to use the non-repetitive scanning characteristics of solid-state lidar to construct the constraint relationship between adjacent point cloud frames,this paper proposes a point cloud feature extraction algorithm based on the combination of curvature and principal component analysis.The feature extraction algorithm based on curvature is used to extract the corner feature points of each laser line,and then the preprocessing point cloud is analyzed by principal component analysis,and the feature points of the surface and volume are extracted according to the discriminant conditions of local geometric characteristics.After the point cloud features are extracted,the feature point cloud of each frame is matched with the local feature map,and the error function is constructed according to the distance from the point to the edge line and the point to the surface.The pose transformation of the point is obtained by solving the nonlinear least squares problem by the optimization method.Thirdly,in order to obtain real-time and robust unmanned vehicle positioning information,this paper adopts a keyframe-based sliding window optimization algorithm to fuse solid-state lidar and IMU measurements in a tightly coupled manner.The algorithm adopts the marginalization method to convert the constraint information outside the sliding window into a priori factor,and uses the IMU pre-integration model to define the IMU preintegration factor between key frames.A lidar odometry factor between keyframes is then constructed using a lidar odometry-based algorithm.The real-time motion estimation of the unmanned vehicle is obtained by jointly optimizing the constraint factors.Finally,in order to verify the accuracy,real-time performance,and robustness of the algorithm proposed in this paper,the public data set and the self-built experimental platform are used to test,and the accuracy and effectiveness of the system are verified.It provides a certain reference value for the research of positioning system based on the fusion of lowcost solid-state lidar and IMU.