Research On Mobile Robot Navigation Method Based On Graph Optimization SLAM

Mobile service robot is one kind of intelligent robots,which has emerged in the fields of family sweeping,restaurant delivery,logistics and transportation recently,and has wide market potential.Autonomous navigation is a necessary capacity for mobile service robots and is the basis of other complex and advanced tasks.SLAM(Localization and mapping simultaneously)and path planning are two key technologies in the navigation system,which have become the hot area of research in academia and industry at home and abroad.In our robot navigation system,the software frame is based on the ROS(Robot Operating System),and the main hardware platform is mainly based on the laser and vision sensor.In order to improve the autonomous navigation capability of the mobile robot,the related navigation methods are studied.The mobile robot navigation system is divided into three layers: human-computer interaction,navigation module and motion control module,according to the function module.In the whole navigation system,the related models about differential wheel,laser and camera are built.The 2D codes about ARToolkitPlus are identified,and the relative poses between the camera and the 2D code are obtained by RPP algorithm.Position error and angular error are measured by some experiments.For the difficulties of detecting the closed loop in a monotonous scenario like a long and narrow corridor,a SLAM method about laser based on graph optimization,which fuses the 2D code information,is proposed.In this thesis,the odometer data is used as the initial pose estimation and then scan matching is used to establish the weak constraint edge in the graph structure.And the 2D code information is used as the closed loop detection signal to establish the strong constraint edge in the graph structure.For the back-end in graph optimization frame,the optimization method of iteration is based on Gauss-Newton or Levenberg-Marquardt to get the optimal configuration in the graph structure.The maps about the office in a company and the corridor in an experimental building are built by using the robot prototype and Turtlebot respectively,and both the two experimental results show that the maps built by adding the 2D code have good consistency.A single static occupied grid map cannot be used for dynamic obstacle avoidance,so the total cost map based on the hierarchical map is adopted actually when planning path.For the shortcomings of the traditional A* algorithm about high searching time,the improved A* algorithm considers the tradeoff between time cost and path length,which is used as the global path planning method of the system.The basic principle of DWA local path planning is analyzed,and the measures to eliminate the abnormal situation like dynamic obstacle are given.During the navigation experiments in the scenario of the corridor,the 2D codes are used as the initial pose estimation and navigation goals.The robot can avoid the dynamic obstacles and finish the navigation tasks successfully.