Research On Robot Hand-eye Collaboration Technology Based On RGB-D Camera

Traditional robots are often confined to structured scenarios due to a lack of perception,but with the development of machine vision,robots are becoming more intelligent and’anthropomorphic’.The key to this is the development of hand-eye coordination technology,which enables the robot arm and camera to work in an integrated manner and includes technical difficulties such as hand-eye calibration and visual servoing.In this paper,we study and explore the robot hand-eye coordination technology based on RGB-D cameras from the perspective of improving the robustness and intelligence of robot hand-eye coordination:(1)Research on hand-eye calibration algorithms based on 3D position information.The current hand-eye calibration algorithm often needs to measure pose information and is affected by the geometric parameters and joint angles of the robot body.In this paper,the rotation part of the hand-eye relationship is corrected by using the axis property to fit the joint axis of the robot arm based on the Kronecker product method using a single joint rotation.Subsequently,a hand-eye calibration method based on the robot arm joint axis and 3D position information is also designed based on the correction algorithm.The method has two solution methods depending on whether the relative position relationship between the robot arm and the calibrated part is solved.The final simulation and experimental results show that the modified method can effectively improve the rotational accuracy of the hand-eye calibration compared to the traditional Kronecker product method,while the method based on the joint axis of the robot arm can improve both the rotational and translational accuracy.(2)Research on hand-eye self-calibration method.In order to improve the robustness and intelligence of the robot’s hand-eye coordination,the hand-eye self-calibration method is studied.The method uses the robot’s operating object as the calibration piece,firstly,the robot arm is controlled to move within a small range to collect data for self-checking the hand-eye relationship and obtaining the reference hand-eye relationship,then the Monte Carlo improvement method is used to combine the camera field of view and the reference hand-eye relationship to obtain the data collection space,and finally the robot arm is controlled to move in the data collection space to complete the collection of calibration data,thus completing the task of hand-eye self-calibration.The simulation results show that all 50 self-calibrations were successfully completed,and the data acquisition success rate was 75.64%.(3)Position vision servo based on online supervision.During prolonged operation or in the event of collisions and severe jitter,the robot’s internal parameters may change,resulting in a degradation of servo performance.When the error of the self-test parameter exceeds a set minimum threshold,the supervisory mechanism will recalibrate the robot’s internal hand-eye relationship parameters to converge the error,and when the error fails to converge or exceeds a set maximum When the error fails to converge or exceeds the set maximum threshold,the robot will issue a warning to ensure the safe operation of the robot.The final simulation and experimental results show that the proposed method has better robustness and safety than the traditional PBVS.