Research On Grasping Technology Of Collaborative Robot Based On ROS

Compared with traditional robots,collaborative robots have comprehensive improvements in perception,decision-making,and learning capabilities,and are widely used in industry,medicine,services,and other fields.As an important capability of collaborative robots,robotic grasping has always been the focus of research by researchers.The grasping process is not intelligent enough,which is still the main obstacle restricting the development of collaborative robot grasping technology.In this paper,combined with the visual recognition technology,the host computer with ROS as the platform is selected to carry out the research on the dynamic target grasping technology of collaborative robots.For the convenience of research,domestic waste sorting is introduced as the specific application scenario of this paper.The research content of this paper mainly includes the following aspects:(1)The D-H parameters of the Sawyer robot were analyzed by the D-H method,the forward kinematics model of the robot was established,and the forward kinematics equation of the Sawyer robot was deduced.The derived forward kinematic equations are simulated and verified by the robot toolbox of Matlab,and the working space of the robot is analyzed by the Monte Carlo method.The robot simulation platform in the ROS environment is built,the robot grasping program is written,and the program is verified and optimized without using a real robot.(2)Two visual recognition schemes are designed: one is a binocular recognition scheme with Kinect V2 as the visual sensor,and the other is a monocular recognition scheme with an ordinary camera as the visual sensor.The principle of camera calibration is introduced and the different camera calibration methods of the two visual recognition schemes in the ROS environment are introduced respectively.A verification method is proposed to verify the accuracy of the calibration results of camera extrinsic parameters.(3)This paper introduces the basic working principle of YOLO V5 target detection,collects more than 10,000 pictures,prepares its own data set,and completes the training of the data set through YOLO V5.Data indicators such as labels and recall rates of the dataset are analyzed,and object detection experiments are carried out to verify the feasibility of the dataset.(4)Inspired by the chameleon grasping tree trunks,a collaborative robot grasping flexible gripper was designed and fabricated using thermoplastic polyurethane as the material.The flexible gripper model was built using SOLIDWORKS software,and the structure was optimized by topology algorithm.The static analysis of the optimized flexible gripper was carried out with the help of the finite element analysis software WELSIM.(5)The communication coordination between target detection and grasping control is completed through the ROS-specific TF tool and topic communication.The sorting experiments are carried out on the monocular recognition scheme and the binocular recognition scheme respectively,and the advantages and disadvantages of each are analyzed.The results of sorting experiments were analyzed,and the success rates of nine categories of garbage recognition were all above 85%,up to 100%,and the sorting success rate was above 85%,up to 94%.At the same time,a comparative grasping experiment between the rigid gripper and the flexible gripper is designed,and the grasping success rate of the flexible gripper for various target objects is better than that of the rigid gripper.