Research On Intelligent Vehicle Mapping Method Based On Lidar And IMU

High definition maps are the most basic and critical part of the process of intelligent vehicles to complete driving tasks.How to construct a high-precision and easy-to-use map has always been the focus and difficulty of intelligent driving research.At present,the application of intelligent vehicles is mainly the automatic driving application of special vehicles in specific scenarios.These scenarios often involve both indoor and outdoor environments,so it is of great significance to build an indoor and outdoor integrated map.In this thesis,for the special scenes,a method of intelligent vehicle map construction based on the fusion of lidar and inertial measurement unit is studied.The research content of this thesis mainly includes the following aspects:Aiming at the problems of large cumulative error and poor mapping effect of traditional lidar mapping methods and the deficiencies of various multi-sensor fusion mapping methods,a point cloud map construction method based on the fusion of lidar and IMU is studied.Firstly,the IMU data is introduced into the traditional lidar mapping method to eliminate the lidar motion error and the feature-based method is used for point cloud matching.Then,a joint optimization model is constructed based on the lidar constraints and the IMU constraints,and the vehicle's motion attitude is estimated by solving the optimization model and a point cloud map is constructed according to the estimation results.In this way,the accuracy of mapping is improved and the impact of accumulated errors is reduced.Aiming at the problem that the traditional map format is complex and not suitable for the use of intelligent vehicles,combined with the needs of intelligent vehicles for maps,a map generation method for intelligent vehicles is studied.Firstly,the map format required by the intelligent vehicles in the scene is determined according to the characteristics of the driving scene and the demand for the map of the intelligent vehicles.Then,the feature points in the lidar point cloud are used for rough road structure extraction,and road reference lines,road slope,roll angle and other structural information are extracted according to the obtained road point cloud.Finally,the road parameter information is written into the map file according to the established map format.In this way,a lightweight map for intelligent vehicles with clear interfaces and easy to use is generated.An intelligent vehicle experiment platform is built,and an intelligent vehicle map construction system is designed and developed.Experiments were carried out in actual scenarios.The experimental results showed that the system's functions and performance reached the expected goals,and also verified the effectiveness and real-time of the method in this thesis.The results of simulation and intelligent vehicle experiment show that the point cloud map constructed by the method in this thesis has higher accuracy and robustness.In addition,the map file generated by the method in this thesis can meet the driving needs of intelligent vehicles and has the characteristics of simple structure,small size and easy to use.