Design And Research Of AGV Autonomous Navigation System Based On Laser SLAM

Since entering the 21 st century,with the rapid development of robot technology and artificial intelligence technology,the era of intelligent robot has come.As a modern intelligent transportation equipment in the field of industrial robots,Automated Guided Vehicle(AGV)has been widely used in manufacturing,warehousing and logistics,special industries and other fields,which can effectively improve production efficiency and reduce workers’ labor strength.However,the application and research of AGV in the field of maintenance are still not intelligent enough.In order to study and design an AGV that can adapt to the working environment of railway maintenance,it’s studied in this paper that the autonomous navigation system of AGV based on laser SLAM.Firstly,the hardware system of AGV platform built in this thesis is introduced,in which the key sensor lidar to realize SLAM is introduced in detail,and the software platform robot operating system used by AGV is introduced.The kinematic model of AGV,IMU pose estimation model and image observation model are established.To prepare for the subsequent autonomous navigation research of AGV.Secondly,the basic principle of Simultaneous Localization and Mapping is introduced,and the construction principle of two-dimensional Occupancy Grid Map is analyzed and introduced.Aiming at the current mainstream 2D laser SLAM technology,the Gmapping algorithm based on filtering and Cartographer algorithm based on graph optimization are theoretically analyzed and studied.Experiments are carried out in the actual environment,the actual mapping effects of the two schemes are compared and analyzed.A better SLAM scheme is selected.Then,based on the construction of environmental map,the global localization of AGV based on a priori map is studied.Aiming at the problems of large amount of calculation and low success rate of the traditional AMCL algorithm based on particle filter in large scenes,a multi-layer global localization method combining laser and visual information is proposed.This method adopts the multi-layer positioning mechanism from coarse to fine,which can realize the accurate global localization of AGV.It realizes the coarse to fine positioning by image retrieval,pose optimization and AMCL.Coarse positioning is an image retrieval and positioning method based on deep learning,and the knowledge distillation method is used to lighten the original image retrieval model and reduce the amount of calculation.Pose optimization uses image feature matching and epipolar geometry to further optimize the coarse positioning pose.Fine positioning uses the optimized pose to initialize AMCL algorithm to accelerate particle convergence and realize the global localization of AGV.The effectiveness of the global localization method is verified by experiments,which provides the initial pose for AGV subsequent path planning and autonomous navigation.Finally,the path planning algorithm to realize the autonomous navigation of AGV is studied,and the autonomous navigation of AGV is realized through the open-source navigation framework of ROS.Then,after the autonomous navigation of AGV is completed and close to the working position.the visual servo positioning control is carried out by identifying the artificial features on the target object through the camera,so that the AGV can realize the accurate positioning at the working position,and the simulation test is carried out in Gazebo through VISP platform.Then the physical test is carried out to verify the feasibility of AGV visual servo positioning control.