Research On Localization Of Low-speed Maglev Test Vehicle Based On Visual-inertial SLAM

The maglev vehicle has become a research hotspot in the field of rail transit because of avoiding the contact between vehicle and wheel-rail,and has the advantages of high comfort,low running noise and environmental protection.At present,many maglev vehicle lines are used for commercial operation.Before the maglev line is officially put into operation,the research on the suspension,guidance and operation control of the maglev vehicle is necessary,in which the real-time and continuous vehicle positioning parameters can provide the research basis for the experiment of the maglev vehicle system.Nowadays,locating technology based on computer vision is an autonomous locating method that has attracted extensive attention.Visual-inertial Simultaneous Localization and Mapping(Visual-Inertial SLAM)integrates the information of the camera and IMU,which has high localization accuracy and robustness,and has great application potential in motion estimation.This paper studies the Visual-Inertial scheme of ORB-SLAM3 algorithm,and optimizes the shortcomings of the algorithm applied to the positioning of the low-speed maglev test vehicle,in order to provide some reference for realizing the real-time and continuous positioning of low-speed maglev test vehicle at low cost.The main contents of this paper are as follows:Firstly,the basic theories involved in Visual-Inertial SLAM are analyzed,including the observation model of camera and IMU sensor,the rotation representation method of rigid body motion in three-dimensional space,the principle of visual pose estimation based on feature points,the mathematical modeling of information fusion between camera and IMU,and the system model and state estimation of Visual-Inertial SLAM.Combined with the previous theoretical basis,this paper combs the principle and process of ORB-SLAM3 algorithm and the localization experiment of low-speed maglev test vehicle based on this algorithm.The Visual-Inertial scheme of ORB-SLAM3 algorithm is deeply studied from the aspects of Visual-Inertial SLAM initialization,tracking,local mapping,loop detection and so on,so as to provide a basis for the improvement of the algorithm.Then,the localization experiment is carried out on the low-speed magnetic levitation experimental line.Through the analysis of the experimental results,the adaptability improvement of the current algorithm applied to the low-speed magnetic levitation line is proposed.Aiming at the problems,the improvement scheme is designed,including the establishment of binary visual dictionary and map reuse algorithm based on maglev test line.Specifically,the creation and optimization methods of visual vocabulary are described,and the data format is transformed to establish binary visual vocabulary.Through the performance test of the exclusive scene visual vocabulary,it is verified that it improves the real-time startup of the system and the accuracy of image matching.The algorithm of map reuse includes the analysis of map data,the preservation and loading of map,and the relocation method based on map reuse.Finally,the localization performance of the improved algorithm is verified on indoor and outdoor magnetic levitation experimental lines.The experiments show that the localization results of the improved algorithm are consistent with the actual operation of the magnetic levitation experimental vehicle.In addition,the simulation experiment is also carried out on the public dataset,and the localization accuracy of the algorithm is evaluated through the localization evaluation parameters.The final Absolute Trajectory Error results show that the improved algorithm can also achieve localization with certain accuracy in the operating environment with a longer line and higher speed.The above research work shows that the Visual-Inertial SLAM algorithm can effectively realize the positioning of the low-speed maglev test vehicle,and the position information has certain real-time and continuity,which can provide help for the research of maglev vehicle system.