Research On Robotic Visual Servoing Control Method

Traditional image-based visual servoing(IB VS)technology directly uses the image information to design a proportional controller.Although the traditional method is robust to the calibration errors and uncertainties of the system model,the traditional method still has some shortcomings:(1)since the stability of the traditional method is only guaranteed in the region closed to the desired position,the traditional method can only accomplish short-range visual servoing tasks;(2)due to the coupling of spatial velocity,the traditional method has serious redundant motion for image trajectories in visual servoing tasks with rotating;(3)the image Jacobian matrix may be singular;(4)the parameter uncertainty of visual servoing model;(5)depth values in the image Jacobian matrix are uncertain,and so on.In this paper,we focus on four problems of image trajectories redundancy,image Jacobian matrix singularity,parameter uncertainty of visual servoing model and depth uncertainty in traditional image-based visual servoing system.Firstly,a depth estimation method is introduced to solve the problem of depth uncertainty in image Jacobian matrix.The depth estimation method is applied to the traditional imagebased visual servoing controller.On this basis,the improved controller is used to generate training set data for BP neural network,and then the design of visual servoing controller based on BP neural network is completed.Because the training set data is directly used in the design of visual servoing controller based on BP neural network,the singularity of Jacobian matrix is avoided.Then,in order to suppress redundant motion of image trajectories and compensate for parameter uncertainties of visual servoing model,a novel visual servoing adaptive tracking control method with time-varying performance boundary constraints is proposed in this paper.Based on a generalized restricted potential function,the new control strategy can effectively overcome the adverse effects caused by calibration and depth errors with the help of pre-given time-varying functions as the bound of tracking errors.Eventually,the tracking error gradually decreases within the given boundary.Compared with existing methods,the new one can effectively suppress the redundant motion of the trajectories on the image plane,thus greatly reducing the risk of feature points leaving the horizon.Moreover,the method can accomplish long distance visual servoing tasks.Finally,in order to verify the feasibility and effectiveness of the visual servoing adaptive tracking control method with time-varying performance boundary constraints,this paper uses industrial robot to build an experimental platform for visual servoing experiments.The effectiveness of the proposed controller is verified by experiments in the practical application of visual servoing.