Research On Lidar SLAM Algorithm For Intelligent Vehicles Based On Multi-Sensor Fusion

In recent years,intelligent vehicle technology has been witnessing a rapid development phase.Among them,the Simultaneous Localization and Mapping(SLAM)algorithm has been widely applied in various fields.In this paper,the multi-sensor fusion SLAM algorithm is targeted to carry out using Lidar sensor,IMU(Inertial Measurement Unit)inertial sensor and GPS(Global Positioning System)global positioning sensor simultaneously.The system contains the following modules: point cloud pre-processing module,laser odometry module,key frame odometry module,IMU HF(high frequency)odometry module,loop closure detection module,and backend optimization module.The preprocessing module accomplishes the motion compensation of the point cloud by utilizing the altitude information output from the IMU HF odometry.The ground point cloud is extracted through multi-region ground segmentation algorithm.Based on the eigenvalue-based information analysis method,the point cloud features with different geometric attributes are extracted and the corresponding mathematical model is established to construct the multi-feature laser odometry.The key frame odometer module enhances the data correlation between different sensors by fusing different types of laser point cloud features and raw data from IMU sensors simultaneously through a tightly coupled optimization method based on the manifold space.The IMU HF odometry predicts the continuous attitude using the optimized parameter information and outputs the attitude data of the intelligent vehicle in real time.The back-end optimization module is responsible for completing the global factor graph optimization by adding various information constraints,including: key frame odometry factor,loop closure factor,normal vector consistency factor,and a priori information factor.Meanwhile,in order to improve the accuracy of the system for long time operation,add priori position constraints when GPS signals are favorable.Finally,the global pose is incrementally optimized by jointly maximizing the a posteriori estimation to improve the algorithm’s robustness of operation and consistency of point cloud map reconstruction in different scenarios.In this paper,laser SLAM algorithm based on multi-sensor fusion is designed for the long time and high accuracy localization and navigation tasks of smart cars that cannot be accomplished by a single sensor.It effectively improves the localization accuracy of smart cars driving in different scenes and provides accurate position and attitude information for the planning system of intelligent vehicles.