Homography-based Visual Trajectory Tracking Of Industrial Manipulator With Iterative Learning Control

Industrial robots need to do some repetitive assembly line work:and Vision-based control can improve accuracy. Due to the lack of depth information which cause model parameters uncertainty, traditional visual trajectory tracking based on image may have the problem of system stability and convergence. This paper proposes visual-based iterative open-plus-closed-loop control strategy which can improve the control system dynamic performance and can ensure the accuracy of the trajectory tracking. Focus on the problem of target object running outside vision which caused by the camera installation of eye in the hands, this paper proposes a non-standard iterative learning control method. During the operation, the target object will run out of the camera horizons, after iterations can ensure that the target object has been within the field of view.The main work of the paper is detailed as follows:(1) We use global gray gradient method to get Homography, instead of the method which firstly gets the feature points such as SIFT algorithm, then use image pixels positions to obtain Homography, so we can use all the information of the planar object which can void errors caused by mismatch and the time is 0.3 times of the method of feature points.(2) Focus on the problem of the lack of deep information, this paper proposes Homography-based visual trajectory tracking of industrial manipulator iterative learning control method and proofs the convergence of the system with the method. Compared to the visual tracking control based on image features, Homography describe the relationship between the two pictures, indirectly reflect the geometric relationship between the cameras, so it has more direct relationship with relationship three-dimensional space which simplifies controller design. The error vector is defined directly based on Homography, without decomposing homography. So it can save time and avoid the problem of multiple solutions of Homography decomposition. Iterative learning control scheme, according to previous error message, adjust the next iteration system input, to achieve precise trajectory tracking. The performances are evaluated based on simulation and experiment results.(3) During the operation, the target object will run out of the camera horizons, this paper proposes Homography-based visual trajectory tracking of industrial manipulator a non-standard iterative learning control method. If during the iteration, the target object runs outside, the next iteration can not use the information after running outside which means time inconsistency. Based on this, we propose two options:1 switching controller, no information on the previous iteration, then the controller switches to only feedback controller; 2 higher order Iterative Learning Control, the last iteration information is missing, choose the optimal iterative information of all the before iterations. The simulation and experiments, indicate that this scheme can effectively solve the problem of the target object running outside vision.