Research On Unmanned Vehicle Positioning Based On 3D Laser Inertial Fusion

With the rapid development of science and technology in my country,unmanned vehicle navigation and localization technology has become the most popular research direction.The core technology of autonomous navigation and localization for unmanned vehicles is Simultaneous Localization and Mapping(SLAM)technology.At present,laser SLAM has become the mainstream research direction because of the high accuracy of laser sensors,which are not affected by weather conditions and can work around the clock.Based on the above background,this article mainly uses multi-line lidar sensors,wheeled odometer and IMU to acquire data and integrate them to conduct research on creating high-precision maps and unmanned vehicle localization.The main research contents of this thesis are as follows:First of all,due to the presence of ground noise in the collected data and the excessively large laser point cloud data,which affects the accuracy and speed of mapping,this thesis first preprocesses the laser radar point cloud and uses the processing method of voxel filtering and down sampling.Based on the original geometric structure of the laser point cloud,the number of point clouds is reduced,and the map creation speed can be increased with almost no impact on the mapping accuracy;and the ray-based ground segmentation method and the ground based on ground plane fitting.The segmentation method filters out ground noise,which can effectively improve the accuracy of mapping.Secondly,the principle of the Cartographer algorithm was analyzed in detail,including the construction of the Cartographer’s occupied grid,the updating of submaps,and the branch and bound search process,which finally realized the establishment of the multi-scene high-precision map.In the process of mapping,the problem of environmental degradation appeared,and a multi-sensor fusion mapping method was proposed,and the problem of environmental degradation was solved by adding wheeled odometer information constraints.Next,this article conducts the research on unmanned vehicle localization on the built high-precision map.First,the real vehicle localization experiment is carried out based on the pure localization mode of the cartographer.Because this method requires high CPU performance,the odometer motion model and the likelihood domain model are established to realize the localization method based on Adaptive Monte Carlo Localization without affecting the localization.Under the premise of accuracy,the utilization rate of the CPU is effectively reduced.Finally,based on the laboratory unmanned vehicle platform,the algorithm was programmed to implement.First,through the comparison of the two algorithms for indoor and outdoor ground point cloud segmentation,it was found that the ground segmentation method based on plane fitting can segment the ground point cloud faster and the separation effect is better.And the map created after filtering out the ground point cloud has higher accuracy;through experiments,it is found that after the down sampling process,the map creation speed has been greatly improved with almost no impact on the map creation accuracy;finally,a long corridor scene was carried out The environmental degradation experiment under the following conditions solves the problem of environmental degradation by increasing the constraints of wheel odometer information;finally,the experimental results are displayed through the visualization tool of the Robot Operating System(ROS).