Research On Accurate Tracking And Localization Technology Of Indoor Mobile Robot

The highly intelligent service of indoor mobile robots benefits from the use of robotic au-tonomous positioning algorithms,but the algorithm has a large amount of calculation and is easily affected by the environment,which makes the robot’s autonomous positioning bi-ased.In order to evaluate the accuracy of robot autonomous positioning algorithms,many mobile robot positioning systems based on indoor positioning technology have been pro-posed.Indoor positioning technology relies on rich indoor information to build algorithm models to determine the location of targets.Among various types of information,visual information is more recoganizable than other information.The indoor positioning technol-ogy based on computer vision is closely related to visual information,which is the focus of this study.In this paper,the sweeping robot is taken as the research object,and single or multiple high-resolution industrial camera is used as the acquisition device of visual im-ages.An positioning system of indoor mobile robot is realized by combining tracking and positioning.The main research content of this paper includes:(1)Implementing a tracking and positioning system for indoor mobile robots.The system uses a monocular industrial camera for image acquisition,highlight the positioning features of the robot with special markers,and designs a tracker and a detector to extract image features.In this system,the positioning of the sweeping robot is realized by the pose es-timation algorithm.The system is divided into four modules.The initialization module realizes the acquisition of the initial data of the system and provides the initial data required for positioning.The feature detection module implements tracking and feature extraction of the sweeping robot and the position information of key features is the basic data for posi-tioning.The robot positioning module realizes the estimation of the camera’s pose and the sweeping robot’s position.The visualization module realizes the display of continuous-time positioning results in the form of paths.(2)Implementing a positioning optimization method based on multi-camera.The method solves problems of occlusion and lack of depth information when the sweeping robot is being positioned by single camera.It builds multi-cameras scene,sets a reference coordinate system to provide a reference object for the determination of cameras’ and robot’s pose,and implements a multi-camera cooperative operation mechanism in a multi-threaded manner to make cameras positioning simultaneously.At the same time,a positioning result fusion algorithm is designed in the method to collates and fuses the positioning results obtained by different times and different cameras to calculate more accurate results of robot position.(3)Implementing a graph optimization method based on the multi-camera scene.The method builds a graph model through the cameras and robot in the scene,solves the op-timization problem of minimizing reprojection errors,achieves overall optimization of the positioning results,and eliminates noise errors accumulated during the positioning.In order to verify the feasibility of the positioning optimization method based on multi-camera,the control experiment was set up between the positioning of single camera and the positioning of multi-camera.In these cases,robots in static and dynamic were posi-tioned separately,and errors in the positioning results were quantitatively analyzed.The experimental results show that the positioning optimization method based on multi-camera reduces the positioning error by about 20%when positioning a static robot and reduces the positioning error by about 34%when positioning a dynamic robot.In order to verify the versatility of the positioning system based on multi-camera,robot positioning tests were conducted in low-brightness scenes and scenes with obstructions.We compare the positioning results of these scenes with normal scenes and analyze the positioning ability of the positioning system based on multi-camera in complex scenes.The results show that low-brightness scenes hardly hinder the positioning of the system.Scenes with obstructions increase the positioning error of the system,but the accuracy of positioning is still considerable.Finally,the overall optimization effect of the graph optimization method on the positioning results is tested,and the feasibility of the graph optimization model in a multi-camera positioning system is verified by comparing the positioning results before and after optimization.