Research On In-vehicle SLAM System Integrating 4D Millimeter-wave Radar And IMU

This thesis studies the vehicle-mounted Simultaneous Localization and Mapping(SLAM)technology based on the fusion of 4D millimeter wave radar and Inertial Measurement Unit(IMU),analyzes the data preprocessing algorithm based on 4D millimeter wave radar and integrated navigation system,studies the positioning and mapping optimization algorithm based on Cartographer,and a vehicle-mounted SLAM system based on the fusion of 4D millimeter wave radar and IMU is concretely implemented.The main research contents are as follows:First,the overall plan of the vehicle-mounted SLAM system is studied.The entire SLAM system mainly includes vehicle-mounted computing platform module,network communication module,video recording module,4D millimeter wave radar perception module,integrated navigation system module,and SLAM algorithm module.According to the needs of the system,the data transmission protocol of the 4D millimeter wave radar and the integrated navigation system module is custom-designed,and the main working mode of the SLAM algorithm module is analyzed in detail.Secondly,the working principle of millimeter wave radar is introduced,a variety of filtering algorithms are discussed,the working principle of IMU is analyzed,and the attitude angle estimation method based on extended Kalman filter is proposed.The multi-sensor fusion algorithm is designed,multiple 4D millimeter wave radar data is spatially fused,multiple sensor data is temporally fused,and a customized fusion data transmission protocol is designed.Thirdly,the point cloud matching method based on the normal distribution transformation is discussed,the vehicle odometer estimation method based on the multi-sensor fusion algorithm is proposed,and the vehicle pose state estimation method based on graph optimization is analyzed.The focus is on the local two-dimensional map construction method based on 4D millimeter wave radar,the dynamic target extraction and filtering method based on the Doppler characteristics of 4D millimeter wave radar is used to improve the accuracy of the local map results,and the global SLAM optimization method based on Lazy decision algorithm is introduced to obtain a global two-dimensional map and vehicle body positioning results with higher accuracy and reliability.Finally,the test plan for the vehicle-mounted SLAM system is formulated,combined with the actual measurement situation,a vehicle-mounted SLAM system host computer software containing the SLAM algorithm is designed.The hardware equipment of the vehicle-mounted SLAM system is installed according to the test plan,the vehicle-mounted test is realized,the test result is analyzed,and the practical feasibility of the vehicle-mounted SLAM system is verified.