| In recent years,with the rapid development of artificial intelligence and sensor technology,intelligent vehicles have gradually become a hot topic in current research.Intelligent vehicle can use on-board sensors to the surrounding environment,in turn,fixing the intelligent vehicle itself,in the moment of on-board sensors,camera although the price is low but its vulnerable to the effects of light,in turn,makes building environment map and the follow-up positioning,and 3D lidar has high measurement precision,distance and is not affected by factors of light.The above advantages are suitable for intelligent vehicles.Aiming at the problems of inaccurate positioning information and large positioning drift in the traditional integrated navigation based on global navigation satellite system and inertial measurement unit,this thesis adopts lidar as the main sensor to construct 3D point cloud map so as to realize the self-positioning of intelligent vehicles to solve the above problems.The main research contents of this thesis are as follows:(1)Aiming at the problems of a large amount of data and excessive useless information in the original lidar data,this thesis preprocessed the raw lidar data.First carried out voxel filtering on the lidar data,which could effectively reduce the amount of LIDAR data,and then carried out radius filtering processing to eliminate outliers.Then the Random Sampling Consensus(RANSAC)algorithm is selected to remove the ground point cloud,and finally the point cloud distortion generated in the process of point cloud movement is removed.(2)Aiming at the problem that only using lidar odometer to construct 3D point cloud map in campus context will produce ghosted point cloud map and low accuracy,this thesis built a 3D point cloud mapping framework mainly based on graph optimization algorithm.Firstly,it conducted experimental comparison on the current mainstream point cloud inter-frame registration algorithms.Finally,the Normal distribution Transform(NDT)algorithm is selected as the point cloud registration algorithm in this thesis,and its pose map is constructed with laser odometer.Then,GPS constraints and the improved loop detection constraints are added.Finally,the whole pose map is optimized to obtain the optimized 3D point cloud map.(3)Aiming at the problem that the positioning accuracy of the traditional GPS and IMU fusion localization algorithm is not high when the GPS signal is lost,this thesis designs a 3D localization algorithm fused with IMU.Firstly,the spatial and temporal synchronization among the multi-sensors is carried out,and then the IMU fusion threedimensional positioning algorithm is designed.In the sensor fusion algorithm,the unscented kalman filter algorithm is selected to fuse the point cloud data of the lidar and the IMU data,and finally the positioning information with high accuracy is obtained.(4)Set up a platform of intelligent vehicle,a brief introduction to each sensor,and then collected the data of the different situations on campus,the above graph algorithm and positioning algorithm of building the experimental verification,the final data show that this algorithm can build a more accurate map of 3D point cloud,effectively eliminate the map ghosting,and the fusion localization algorithm can get high localization results. |
