Research On Control Algorithm Of Arm Robot Based On Visual Feedback

In 2015,the State Council issued the action program "Made in China 2025",which pointed out the direction from a large manufacturing country to a strong production country.Industrial robots are the key to realizing intelligent manufacturing and industrial transformation.Arm robot that is a multi-joint manipulator or a machine with multiple degrees of freedom is an important type of robot and plays an important role in industrial field.However,the traditional arm robots lack the ability to sense the environment and make quick adjustment according to real-time changes.This leads to the lack of flexibility and robustness of traditional robots and greatly limits their application in dynamic scenes.Therefore,this paper combines the vision system with the robot system to carry out the research of robot control algorithm based on visual feedback,which is beneficial to the object tracking,grasping,placing and obstacle avoidance.Firstly,the development status of visual servo technology at home and abroad are analyzed,and expounds the research results of visual servo in robot grasping.Secondly,Kinect 2.0 is used as a visual sensor to collect color images and depth images and study the object recognition and pose estimation of textureless objects.For the rule objects,the Euclidean clustering segmentation and RANSAC are used for object recognition.For the general objects,feature matching of poin cloud based on the color segmentation is used for object detection and pose estimation.Then,the related theories of robots are studied to establish the structural model and kinematics model of the SD700E robot producted by STEP.Based on the robot model,the basic motion planning in the joint space and the Cartesian space is studied,and the quintic polynomial is selected as the basic model to construct the motion with continuous jerk.In addition,in order to eliminate the pause between two trajectories,a symmetry-based transition trajectory planning method is adopted to realize the adaptive continuous motion planning of multiple trajectories.For the operation of dynamic objects,a free boundary motion model based on three equal cubic polynomials is proposed.The robot can move from any current pose with any current velocity and acceleration to the any desired pose with any desired velocity and acceleration within the effective motion range.With them,based on Matlab and ROS,the motion control simulation of the robot tracking,grasping and placing for the static object,the uniform linear motion object,the uniform circular motion object and parabolic moving object is completed.Finally,the actual tests of robot control algorithm based on the visual feedback are carried out by the grasping experiments of a rectangular carton.In the vision module,the single detection duration of the object recognition and pose estimation realized by the three-step preprocessing including the effective distance setting,the color segmentation and the downsampling pixel reconstruction is less than 250ms.In the experiments,according to the object pose from visual feedback,the robot can accurately and stably track,grasp and place the static and dynamic objects.The verification results show that the visual detection algorithm used in this paper satisfies the requirements of real-time and stability.The proposed robot motion planning algorithm is suitable for the robot control system based on visual feedback.It has abilities to achieve the adaptive and anti-interference planning for both static and dynamic scenes.