Research On Uncalibrated Visual Servo Control System For Manipulator Based On Improved Q-Learning

With the continuous progress of society,people urgently need more intelligent machines to replace people to complete those dangerous or more complex tasks,among which the use of manipulator is very wide.The uncalibrated visual servo control for manipulator is controlled directly according to the description of the target in the image space,and it does not need to be controlled in the Cartesian space,so as to avoid a series of defects caused by the complex space transformation in the calibrated visual servo control,it can directly designs the control law in the image space,which has become an important research direction.Because the uncalibrated visual servo control for manipulator directly designs the control law in the image space,there is the problem of trajectory optimization in Cartesian space.In addition,a large number of industrial manipulators can not support advanced control systems such as ROS,which makes it difficult for a large number of applications and intelligent algorithms based on ROS to be directly applied.Based on the above problems,this paper adds Q-learning algorithm and ROS to the uncalibrated visual servo control system for manipulator.On the basis of ensuring the stability of the system,it optimizes the trajectory of system for manipulator and improves the speed of the system.The main research contents are as follows:(1)Propose an control method based on uncalibrated visual servo for manipulator of improved Q-Learning.Improved the action selection strategy and state-action value function in the Q-Learning algorithm,so as to reduced the amount of learning tasks,improved the convergence speed of the system,and continuously optimized the motion trajectory of manipulator in Cartesian space through self-learning,and improved the speed of motion control.(2)Proposed a structure of visual servo control system for manipulator based on ORIN2-ROS middleware.Design and adopt ORIN2-ROS middleware,so as to complete the interaction between system for manipulator and ROS,and then construct a complete platform of visual servo control system for manipulator.(3)Built the simulation experiment platform based on MATLAB and the physical experiment platform based on DENSO manipulator,and designed the experimental plan,after comprehensive comparison of various experimental results,verified the effectiveness of the structure and method proposed in this paper.