Research On An Intelligent Control System Of Manipulator Visual Servo

With the rapid development of technology,robots are widely used in industrial production and life services.Vision is a manifestation of robot intelligence.The visual servo of the manipulator has become an important research direction in the field of robots,and the intelligence level of the manipulator has also been paid more and more attention.Introducing intelligent control into the robot's visual servo control system to improve the robot's intelligence level,improve the speed,stability and accuracy of the system is the current research hotspot and focus and has broad application prospects.The thesis simulate the grasping behavior of human visual deviation.Aiming at the grasping process of manipulator,designs a visual servo intelligent control framework for manipulator motion control,and presents a visual servo control method for manipulator based on human simulated intelligent control(HSIC).Aiming at the optimization of the control parameters of the humanoid intelligent controller,a reinforcement learning algorithm is introduced to perform online learning and optimization of the relevant control parameters of the controller to realize the robot's visual servo intelligent control grab task.The main work completed is as follows:(1)The thesis presents a visual servo system control method of Eye-in-Hand manipulator.Analyzed the visual servo system of the manipulator with eyes in hand.Based on the theory of human simulated intelligent control,an improved human-intelligent multi-mode control method is designed.Determine the target feature state,form the dynamic information space and obtain the feature model by dividing the target state deviation and its change rate and other information,and using Kalman filter-based feature state estimation to optimize the output of feature deviations to solve noise interference problems in sports vision;A multi-mode control strategy based on feature model and feature identification is proposed.Multiple control modalities are designed according to different feature states.A reinforcement learning algorithm is introduced to perform online learning and optimization of the control parameters of the control modalities to obtain better control effect.In the Matlab simulation environment,the control algorithm simulation experiment was carried out using the robot toolbox and the visual toolbox,which verified the effectiveness of the human-like intelligent control method based on reinforcement learning in this paper,and obtained better control effects.(2)The thesis designs and implements the robot visual servo experiment platform based on the robot operating system(ROS),and carries out the robot gripping experiment.According to the requirements of the control system,designed the control framework and hardware architecture of the robot visual servo control system.The manipulator control system uses the UR5 cooperative manipulator,Robotiq-85 gripper and MIC-7700 industrial control computer,and uses the D435 i camera to realize eye-in-hand visual perception.Based on the above system,set up a system experimental platform,and build a manipulator visual servo system capture experimen.The results of the experiment verify the speed and accuracy of the method and system proposed in this paper.