Vision And Odometer Fusion Positioning Algorithm In Closed Environment

With the continuous development of sensors,computer technology and manufacturing,the development of mobile robots has reached a climax.In particular,indoor mobile robots,such as the ubiquitous restaurant service robots,enter the sweeping robots of thousands of households,and contribute to the logistics industry,such as Automated Guided Vehicles.However,in order to complete the prescribed tasks,it is inseparable from precise positioning technology.The traditional electromagnetic guiding and positioning technology is complicated in arrangement and is easily interfered by metal objects,and the long-term positioning accuracy of the odometer is relatively low.Therefore,this paper proposes a method of locating visual information and odometer information to solve the problems of poor long-term positioning accuracy and complex positioning process.The main work of this paper is as follows:(1)Before merging the vision and odometer,it is necessary to calibrate the relative positional relationship between the sensors.At present,the existing calibration methods have insufficient calibration accuracy,rely on artificial landmarks,and can only be off-line calibration.This paper proposes an algorithm that does not depend on any artificial landmarks and can be calibrated in real time.The calibration results obtained by the traditional online calibration algorithm are often linear solutions,and the results are often six degrees of freedom.Therefore,this paper proposes a parameter joint optimization method to improve the calibration accuracy,including: camera pose,camera internal parameters,three-dimensional coordinate points,and external parameters of the camera and odometer.The robot moving on the plane,because the five degrees of freedom between the body and the fixed camera is considerable,it is solved by the step method.(2)This paper optimizes the fusion and positioning problem of vision and odometer using graph-based optimization method.Compared with the single vision localization algorithm,the fusion localization algorithm not only adds the reprojection error constraint of the image,but also adds the pose constraint between the odometer and the robot motion plane constraint.Through experimental verification,the re-projection error constraint of the image,the pose constraint of the odometer and the plane constraint of the robot are jointly optimized,which can improve the positioning accuracy and robustness.(3)The cumulative error of the robot during long-term motion will result in poor map consistency.In response to this problem,the closed-loop detection method based on visual features is used to correct the error and spread the error throughout the motion.In addition,the system proposed in this paper is equipped with real scale information,so when the pose is optimized for the whole map,the pose representation uses the special Euclidean transformation(SE3).Finally,the simulation experiment and real environment test prove that the proposed fusion location algorithm is better in accuracy and robustness.It has research significance and application value for the positioning problem of mobile robots in closed environments such as indoor and underground parking lots.