| In today's world,with the progress of science and technology,robots have begun to appear in our field of vision,and serve various fields,replacing or assisting humans to complete various tasks.Robots have become indispensable in many aspects,becoming human right and left arms,and are used in various fields.Therefore,in the face of the growing demand for robot navigation and positioning,in order to solve the problems of robot positioning in indoor,deep sea,or high-rise metropolises,this paper expands the loose combination of Xtion sensors and inertial measurement unit(IMU)Research on indoor positioning algorithms for mobile robots.Visual odometry(Visual Odometry,VO)means that when the camera moves in an environment with sufficient lighting and clear texture,it gradually iteratively obtains its own position and posture by photographing the changing characteristics of the surrounding environment.The camera is generally mounted on a carrier.The carrier mentioned here may be a robot or a vehicle.The Xtion visual odometer used in this article is an RGBD camera,which can collect two kinds of information at the same time during the movement of the carrier:color information and depth information,which is different from the traditional monocular binocular visual odometer.During the movement of the robot,the Xtion camera can collect a point cloud with a higher density than the traditional visual odometer.In this way,by using image feature extraction and matching algorithms,a more accurate point cloud matching pair can be obtained,and at the same time,some can be reduced.The wrong point cloud matching pair,the resulting robot trajectory will be more accurate than the monocular binocular odometer.We know that there are some prerequisites for obtaining a precise robot movement trajectory through a visual odometer.For example,as mentioned above,the ambient light source is sufficient,and the surrounding environment has enough clear textures.If the scene where the robot is located is dark,the light is insufficient,or there are multiple glass scenes in the surrounding environment,and the texture is unclear,it will lead to a reduction in the efficiency of image feature matching,and at the same time,incorrect matching will increase,affecting the accuracy of the final result.This requires further improvement and improvement of existing image feature matching algorithms in future research.It is well known that when using a visual odometer for navigation and positioning,the camera equipped with it will collect environmental information of several Gs or even tens of Gs,which will put forward higher performance on existing feature processing algorithms and computer performance.Requirements,and the reality is that the processing efficiency of existing algorithms cannot keep up,so that the positioning time of the robot will increase.Obviously this does not meet the requirements of fast and accurate positioning.Looking back at other existing navigation and positioning sensors,we will naturally think of the inertial measurement unit.The IMU has the characteristics of fast,simple and completely autonomous navigation,which can make up for the lack of Xtion in real-time and accuracy,but the inertial measurement unit is positioned There is a problem of accumulation of displacement errors,and it cannot be accurately positioned for a long time.In this paper,the combined navigation and adaptive fading extended Kalman filter(AFEKF)method is used to fuse visual SLAM positioning and inertial measurement unit data to improve accuracy and innovation.The image of the target frame and the reference frame obtained by using a visual sensor,and the subsequent processing of the image of the target frame and the reference frame are processed by a series of algorithms to obtain reliable point cloud information,and finally the movement trajectory,and then combined with the IMU.Finally,multiple sets of different trajectory combined navigation and positioning experiments are carried out indoors to verify the algorithm. |
PDF Full Text Request