Research On Slam And Navigation Based On Multi-sensor Fusion

Simultaneous localization and mapping(SLAM)is a process in which a mobile robot uses its own sensors to determine its own position and surrounding environment.This technology is the premise of realizing autonomous navigation of an intelligent mobile robot.At present,the use scenarios of mobile robots are becoming more and more complex,which makes SLAM algorithm based on single sensor difficult to run robustly in complex environment.For example,the vision camera will fail to track in scenarios with insufficient illumination,missing features,and interference from moving objects;2D lidar can only obtain one plane of information can not completely restore the scene information;Inertial measurement unit(IMU)has zero offset,and the use of integral method to solve the pose will produce cumulative error.Therefore,how to efficiently and robustly complete the target tasks of mapping and navigation in complex scenarios is an urgent problem to be solved.According to the characteristics of different sensors,multi-sensor fusion SLAM has become a development trend.This thesis aims to improve the accuracy and robustness of mobile robot positioning and navigation in the indoor complex environment.Therefore,the portable sensors commonly used by mobile robots are studied.By studying the sensor model and its own characteristics,and summarizing the advantages and disadvantages of each sensor,the multi-sensor information fusion of visual camera,IMU,lidar and other sensors is studied to achieve more accurate and robust SLAM and autonomous navigation for mobile robots in unknown environments.The main work of this thesis is as follows:First,the models and measurement principles of the three main sensors mounted on the mobile robot platform are studied,including the pinhole imaging model and distortion model of the vision camera,the error and motion model of the IMU,and the ranging principle and observation and noise model of the lidar.Second,researched and designed a positioning algorithm based on non-linear optimization with tightly coupled vision and IMU,and established visual-inertial odometry(VIO).The introduction of high-frequency IMU data information based on the pure visual SLAM algorithm improves the robustness of the system when responding to rapid motion changes.At the same time,in order to solve the problem that the visual SLAM cannot operate stably under the condition of insufficient illumination and lack of features,and the map established by the visual SLAM algorithm cannot be used for navigation,a fusion lidar positioning mapping module is proposed,which can assist in the case of visual feature tracking failure Locate and build a two-dimensional grid map to achieve navigation tasks.Finally,the navigation framework of the mobile platform is proposed.Through the study of path planning algorithms,it is proposed to use a combination of global planning and local planning for path planning.Design and build a mobile experimental platform to realize the information interaction between the mobile platform's perception layer,function layer and hardware layer.Use this platform to conduct multiple experiments on the subject proposed in this paper to verify its effectiveness and feasibility.Experiments show that SLAM combined with visual inertia and lidar can effectively improve the positioning accuracy,and the root mean square error of its absolute trajectory error is about 0.03 m tested under standard data.Under multi-sensor fusion,a more reliable 2D grid map is constructed to perform navigation tasks,so that the robot can reach the navigation destination smoothly and accurately,and the robustness of the positioning and navigation of the mobile platform is improved.At the same time,a three-dimensional point cloud map of the indoor environment can be built to obtain more environmental information for the exploration and measurement of unknown environments.