| In the recent years,large freight train marshalling stations such as Zhengzhou North Railway Station have basically realized automation.But the unhooking operation on the hump station still relies on manpower.Meanwhile,the unhooking work is highly intensity,pressure and dangerous.This manual method cannot adapt to the development of railway transportation in our country because of the large freight volume and fast growth speed.With the advancement of science and technology,the improvement of railway transportation automation has been put on the agenda,and the automation of train unhooking has become a burning problem.Based on the research on the operation of the six-degree-of-freedom manipulator at the hump unhooking station,the engineering problem of the automation of the unhooking station at the marshalling station is studied and solved.The specific research content is as follows:1.The design of the unhooking mechanical arm system and the establishment of the kinematics model is completed.The manipulator hardware system is settled by hardware selection and designing the six-degree-of-freedom manipulator model.The unhooking manipulator software platform is built under the Robot Operation System(ROS).The interface program between the upper computer controller and the lower computer is designed to realize the control between the upper and lower one.Finally,the experiment complete the forward and inverse kinematics analysis of the manipulator in the workspace through kinematic modeling and perform simulation verification on the forward and inverse kinematics in MoveIt!;2.The recognition of train coupler handles based on deep learning is realized.The recognition effect of the three network models—SSD-MobileNetvl,YOLOv3-Darknet53,YOLOv3-Tiny—on the coupler handle is compared from three aspects such as recognition accuracy,detection speed,embedded deployment capability,etc.According to comparison,this paper selects the the deep learning model,training weight file,and ensure the weight file can be embedded in ROS;3.The research is to improve the double S-type speed interpolation curve.Specifically,the paper studies the cubic polynomial interpolation and fifth-degree polynomial interpolation algorithms used commonly is to improve the double S-shaped speed interpolation curve,and plan a smoother acceleration curve and speed curve.The improved double S-shaped speed interpolation curve is used to realize the end effector of the robot arm in the flute Carl space moves with straight trajectories and arc trajectories.The characteristics of the RRT algorithm are studied and the obstacle avoidance effect of the algorithm is verified through simulation.4.The last task of this paper is to Realize the recognition and positioning of train coupler handles in the ROS framework.The tester calibrates the internal parameters of the camera to eliminate image distortion,and performs "hand-eye calibration" on the camera and the robotic arm to obtain the matrix transformation relationship between the camera and the robotic arm.The left and right cameras are used to identify the train coupler handles.Meanwhile,the binocular distance measurement principle is used to obtain the vertical distance between the handle and the camera coordinate system under the circumstance of the recognition on coupler handles,to calculate the three-dimensional coordinates of the train handle in the robot arm coordinate system.The simulation environment is built in the Gazebo simulation software and the unhooking process of the unhooking robot arm is simulated. |
