Robot Autonomous Navigation Algorithm Based On Vision And Inertial Sensor

With the development of times,the demand for indoor mobile robots is increasing,and the requirements for navigation algorithms are also increasing.Navigation algorithms include positioning,mapping,planning,obstacle avoidance,motion control and so on.And positioning is the core of the navigation process..There are many options for indoor positioning,but GPS-based positioning is not effective in indoor.The odometer rule based on encoder needs to specify the initial position of the robot,and the error will accumulate as the distance increases.In recent years,Simultaneous Localization and Mapping(SLAM)technology based on vision has become more and more popular.However,pure visual SLAM also has some shortcomings,such as poor performance in weak texture regions or dynamic environments.Therefore,this paper studies and improves the visual SLAM scheme,and adds some additional sensors to complete an autonomous navigation scheme for indoor mobile robots.The main work is as follows:(1)The principle of positioning mapping system is studied.Investigate and survey the development status of visual SLAM and visual inertial SLAM.The theoretical problems involved in SLAM system,such as camera model,pose calculation and global optimization,are deduced mathematically.(2)Visual inertial odometry is investigated.The visual odometer is improved by using the encoder information and gyroscope information provided by the hardware system.The ORB feature is extracted from the image,and the feature points are matched according to the prior information provided by the encoder and gyroscope.Then,according to the matched feature points,the PnP method is used to minimize the reprojection error and restore the relative motion of the camera.(3)Loop detection and optimization are carried out.The loop detection based on word bag model is applied to SLAM system,and research the IMU pre-integration model.At the same time,the results of loop detection and IMU information are added to the global pose graph optimization to further optimize the accuracy of localization and mapping.(4)Navigation system is built.Based on SLAM algorithm,path,planning,dynamic obstacle avoidance,motion control and other algorithms are studied and improved to build the navigation system.(5)Build a hardware platform.Transplant related algorithms,implement its software,and finally experimental verification is carried out,so as to make the accuracy of the whole navigation system less than 5% and stable operation in complex indoor environment,ultimately achieve the practical purpose..