Research On Robot Localization Based On Sparse Direct Visual SLAM And Inertial Navigation

Mobile robot positioning technology is one of the key research topics in the field of navigation and positioning,and is also the basis for robots to achieve autonomous navigation.Since the vision sensor can obtain scene information in real time,and the information provided does not depend on prior knowledge,the vision sensor has a strong ability to adapt to the environment and high positioning accuracy,and is widely used for robot positioning.However,when the robot moves too fast or is blocked,the visual sensor has the problems of tracking loss and invalid data.The inertial sensor is stable in the short-term prediction of fast motion,but there is a problem of accumulated drift that increases with time.In response to these problems,by fusing and estimating the two data,fully combining the advantages of both,combining the high precision of visual SLAM with the stability of inertial navigation can effectively improve the positioning accuracy of the robot.The main research content of this paper is to integrate inertial and visual information to solve the problem of precise positioning of mobile robots.The specific research content is as follows:1.Researched the inertial navigation system.The content includes the classification of inertial navigation systems,positioning principles,and the transformation relationship between the coordinate systems used in the navigation solution.First,the two major classifications of inertial navigation systems are introduced,and then the positioning principles of inertial navigation are explained,and finally the derivation Attitude,speed and position update algorithm.2.The Sparse direct visual SLAM positioning algorithm is studied.First,the three categories of direct visual positioning system and the mathematical model of sparse direct visual SLAM positioning are explained.Secondly,the sparse direct visual SLAM has a gray-level invariance assumption,which causes the sparse direct visual SLAM to be greatly affected by light changes.,Proposes an online camera photometric parameter estimation algorithm based on ORB feature points,and uses the block photometric parameter estimation based on it,and combines it with the sparse direct method visual SLAM to reduce its exposure to light changes.It can enhance the robustness of illumination and improve the positioning accuracy.Finally,the feasibility of the algorithm is verified through design simulation experiments.3.Researched the visual-inertial combined positioning algorithm.First,derive the IMU pre-integration expression.Secondly,initialize the parameters of the visual-inertial combined positioning system.Then,according to all the optimization variables in the objective function,the residual term and the derivative of the objective function with respect to the optimization variables are derived.Finally,a simulation experiment is designed,and the EUROC data set is used for testing to verify the effectiveness and accuracy.4.Set up a visual-inertial combined positioning system simulation experiment platform on a personal computer.The software test platform is composed of ROS,and the common functions in ROS are introduced.The hardware test platform is composed of Voyager ? autonomous mobile robot,MYNT binocular camera and K708 The high-precision positioning GNSS board is composed of a simulation experiment by using the Voyager ? autonomous mobile robot equipped with a small Mi binocular camera to collect data.The experimental results verify the effectiveness and accuracy of the visual-inertial combined positioning algorithm proposed in this paper.