Research On Uncalibrated Visual Servoing Control System Of Electronic Component Assembly Robot

At the current 3C electronics and semiconductor industry,large-scale robotic automated assembly operations have become a reality.With the breakthrough of machine vision technology,robot visual servoing control has been widely applied in the assembly of electronic components.However,the operating model of classical robot visual servoing relies on regular calibration of the device,so as to run on the preset trajectory and meet the accuracy requirement.In complex industrial environment,calibration would become more complicated and highly frequent,and the accuracy and robustness of calibration are poor.Robotic uncalibrated visual servoing technology eliminates the tedious calibration process,has more advantages in control efficiency,precision and robustness,and further improves the intelligence of robot electronic components assembly.Based on the image-based uncalibrated visual servoing control,this paper designs the uncalibrated visual servoing system for electronic component assembly,the main research contents include:(1)The improved RCF algorithm based on deep learning technology is used to extract the edge of electronic component.The CBAM attention mechanism module is used into the convolutional layer of main network of the RCF model,and the cascaded form network and feature fusion module are added into the side output network to fuse the side output of each layer with the feature output of the previous layer,so that the high-level information is transmitted to the low-level feature.The improved RCF algorithm reduces the interference of internal and background texture information of electronic component,makes the obtained edge of electronic components more fine,image feature data more accurate,and improves the robustness of uncalibrated visual servoing system.After image process and feature extraction,the line feature information required by the objective function of the uncalibrated visual servoing control system is obtained.(2)Online estimation of Jacobian matrix based on improved robust information filter algorithm.The robust information filter sets the errors generated by the linear approximation of Jacobian matrix as random bounded noise,then realizes accurate estimation of Jacobian matrix in the unknown system noise environment,but the algorithm does not consider the outliers in the measurement data.The improved robust information filter algorithm,with the introduction of adaptive residual weight filtering algorithm,significantly suppress the outlier data in the observed data,and achieve the effective estimation of Jacobian matrix.The improved algorithm has better robustness to industrial production environment.(3)Establishing the experiment platform of uncalibrated visual servoing system for electronic component assembly robot.Developing the HMI of manipulator based on Py Qt5,Raspberry Pi 4B is used as the system controller,Arduino board is used as the controller of the manipulator,industrial camera is used to obtain image feature information,then writing the control program,designing the grabbing experiment of electronic components,finally verifying the validity of uncalibrated visual servoing prototype system for electronic components.