Research On Uncalibrated Visual Servo Positioning Control Method For Riveting Robot

As the core of the automatic riveting system,riveting robot is widely used in the assembly of parts with high requirements of size and precision in the fields of aerospace,rail transportation and other industries,its performance directly affects the machining efficiency and quality of the whole manufacturing system,so it is important to study the high-efficient,flexible,stable and reliable robot riveting system for industrial assembly.In this paper,the problem of large chattering and low tracking precision caused by external disturbance in the process of riveting is studied by constructing a non-calibrated visual servo positioning system for a riveting robot,in order to solve the problem of poor riveting quality caused by the difficult positioning of multi-nail riveted workpieces with small clearance,the corresponding intelligent control algorithms are designed.The main research work of this thesis is as follows:1.In order to solve the problem of large chattering and low tracking precision caused by external disturbance in riveting process,an adaptive control method combining disturbance observer and fractional sliding mode controller is designed,the chattering in robot motion is eliminated,and the trajectory tracking accuracy is improved.Firstly,the local dynamic model of the riveting robot is established by introducing the time delay estimation strategy,and the disturbance observer is used to observe the known external disturbances in real time.Secondly,for the chattering problem,the fractional sliding mode surface is designed to replace the traditional sliding mode control,and a new reaching law is adopted to make the system move continuously and smoothly on the switching surface,aiming at the unknown interference outside the system,an adaptive strategy is designed to completely compensate the external interference.The experimental results show that,compared with fractional sliding mode control,the proposed method can effectively eliminate the chattering problem of the system and improve the trajectory tracking accuracy of the riveting robot,the peak value of the tracking error of the first three joints of the robot is reduced by 50%,59% and 63%respectively,which ensures the accurate control of the robot during the riveting operation..2.In order to solve the problem of poor riveting quality due to the difficulty of locating multi-nail riveted workpieces,a method of combining robust Kalman filter and neural network is proposed to realize the automatic positioning of multi-nail riveted workpiece.Firstly,the vertex of the multi-nail riveted workpieces is extracted as the characteristic point of the uncalibrated visual servo system.Secondly,the robust Kalman filter is used to predict the image Jacobian matrix in real time,and the estimation error of the system parameters is compensated dynamically by the neural network on-line.The problem of low recognition accuracy of the image Jacobian matrix caused by the correlation of unknown system noise and the delay of "image space-operation space" is solved.Finally,a depth estimator is designed using the least squares method to estimate the depth value in the workpiece positioning by changing data of the robot’s motion state and image characteristics.The simulation results show that,compared with the traditional Kalman filter estimation method,the convergence time of the end speed of the robot is reduced by 30%,the absolute value of image characteristic error is less than 1 pixel,the trajectory is stable,and there is no oscillation feedback phenomenon.This method can ensure that the robot can quickly complete the precise positioning task of multi-riveted workpieces.To address the problem of poor riveting quality due to difficulties in positioning and alignment of multi-nail riveted workpieces,a combined robust Kalman filter and feedforward neural network approach is proposed to enable automatic positioning and alignment of multi-nail riveted workpieces by feature points.First,the vertices of the riveted workpiece are extracted as the feature points of the uncalibrated visual servo system.Secondly,the robust Kalman filter is used to predict the image Jacobian matrix online in real time,and the estimation error of system parameters is compensated dynamically by the neural network online,which solves the problem of low accuracy of image Jacobian matrix recognition caused by the mutual correlation of unknown system noise and the delay of "image space-operation space".Finally,the depth estimator is designed by the least squares method,and the depth value is estimated by the robot’s motion state and the change of image features to obtain the depth of the motherboard that needs to drop after the workpiece positioning alignment.The simulation experimental results show that the method reduces the robot end speed convergence time by 30%,the absolute value of image feature error is less,the running trajectory is smooth,and there is no oscillation feedback compared with the existing estimation method using Kalman filter,which can ensure that the robot can quickly complete the positioning alignment task of multi-nail riveted workpiece.