The Visual Tracking And Control For A Train Uncoupling Robot

The uncoupling work in the station is continuing,sometimes the work efficiency is difficult to guarantee and the phenomenon of picking the wrong hook will happen if the uncoupling work is done by people,so it is necessary to design a train uncoupling robot to finish the uncoupling work to ensure the work efficiency and pick the right hook.In order to finish the uncoupling work,the train uncoupling robot not only should have the ability to identify the target and track the target steadily,but also have the ability of planning and moving for arriving the expected position.The prerequisite of finishing the uncoupling work is to design the train uncoupling robot system from all sides.This paper mainly analyses and studies the walking mechanism of the train uncoupling robot,the walking mechanism is an important part of the train uncoupling robot.The walking mechanism not only should ensure that the robot can catch up with the hook accurately,but also ensure that the robot keeps moving at the same speed with the hook after catching up with the hook.According to the motion requirements of the uncoupling robot,the main work of this paper is as follows:First of all,give a brief introduce of the basic structure and working principle of the train uncoupling robot,it can be seen from the working principle that the robot needs to identify the hook and know the specific position of the hook before starting.High real-time SIFT algorithm is used to identify the hook,SIFT algorithm can identify the target accurately although there are changes in size or direction of the target.Secondly,in view of the situation that the size of the moving target will change in the image,design a tracking algorithm which can adjust the tracking window adaptively according to the size change of the target.The improved Mean Shift algorithm is used as a moving target tracking algorithm,the tracking window can adjust automatically when the size of the target is bigger in the image.The Kalman filtering algorithm is also used in target tracking to solve the situation that the tracking window will be lost when the part of the target is temporarily blocked.Again,fuzzy inference rules,which can make decisions according to human thinking,are used to plan the moving trajectory of the robot because the target is in constant movement.Planning the displacement,velocity and acceleration of the robot can ensure that the train uncoupling robot can move to expected position smoothly.Finally,generalized predictive control algorithm is used to realize the high precision control of the robot moving trajectory,and Elman neural network algorithm is used to get the predictive model of predictive controller,which can solve the difficult problem of nonlinear system modeling.