Localization And Mapping For HTS Maglev Test Vehicle Based On VIO

As a novel mode of transportation,maglev can overcome the friction between wheels and rails and run fast above the track.Compared with traditional maglev,high-temperature superconducting(HTS)maglev has great potential for development,because it’s able to realize self-stable levitation without active control.In navigation,mapping and odometry for indoor and outdoor environment,Visual Inertial Odometry(VIO)is the computational problem that constructs a map of an unknown environment and keeping track of real-time location simultaneously by camera and Inertial Measurement Unit.In this process,the camera can obtain rich environmental characteristic information,and effectively reduce the IMU drift problem;the IMU can collect the acceleration and angular velocity of the carrier at high frequency,and obtain the real-time velocity through integration operation,thereby reducing the impact of fast motion on the observation data.In order to explore the accuracy and robustness of the VIO method under the HTS maglev test system,this paper focus on the different methods between VI ORBSLAM and VINS-Fusion in the visual-inertial alignment,and absorbs the advantages of the two to improve the positioning accuracy of the VINS-Fusion.Through experimental analysis of the Eu Roc dataset in different scenarios,compared with VINS-Fusion,the accuracy of the improved VIO method is improved.On this basis,verify the feasibility of the VIO in measurement environment.The main work of this paper includes the followings:Firstly,the relevant theories related to the data fusion problem of the camera and IMU sensor are elaborated and derived.The IMU dynamic equations and incremental error equations based on the median method are studied,and the process of data fusion based on the tightly coupled VIO method for low-frequency camera and high-frequency IMU data is summarized.Secondly,the applicability of classical tightly coupled VIO methods such as OKVIS,VINS and VI ORBSLAM in the measurement environment are analyzed.Through the preliminary analysis results,the differences between VIO SLAM and VINS-Fusion are expounded from five aspects: system framework,image information extraction,keyframe selection,positioning accuracy and system robustness.Although VINS-Fusion is slightly inferior to VI ORBSLAM in positioning accuracy,it is more robust in fast motion.Therefore,VINS-Fusion is more suitable for HTS maglev ring line.In addition,for the reason that the absolute trajectory error(ATE)of the system initialization phase of VINS-Fusion is large,the visual-inertia alignment process of VINS-Fusion is explored.Hereafter,VINS-Fusion directly integrated the error caused by the accelerometer bias into the gravity.Within the sliding window,the acceleration of gravity was estimated by minimizing the error between the IMU preintegration increment and the estimated value.In response to this problem,VI ORBSLAM’s visual-inertial alignment method was transplanted to the VINS-Fusion initialization module.The EuRoc dataset verifies that the improved method involves variables from visual-inertial alignment is convergence and the positioning accuracy is also improved.Finally,the S1030-IR-120/Mono camera was used to verify the suitability of the improved VIO under the HTS maglev ring line on ROS Kinetic platform.Experimental results show that the improved VIO algorithm is suitable for the environment with repeated textures,and has certain development prospects and research value.