| In recent years,the techniques of machine vision and other correlation domain has been deeply developed.Visual servoing control of manipulators with image information can enhance the diversity of information acquisition,expand the spatial cognition and adaptability,and improve the capacity for accurate recognition and delicate operation.In addition to industrial and medical applications,manipulator visual servoing has been expanded to more complex working space,such as deep-sea exploration,explosive ordnance disposal,disaster exploration,etc.The traditional industrial manipulator performs success in assigned servoing tasks.However,manipulator cannot make effective decisions due to the change of working environments.Thus,manipulators should be master in extracting accurate image information and making feasible motion decisions,which is also a subject for positioning and tracking control of manipulator visual servoing systems.Therefore,the techniques of the manipulator with active positioning and tracking ability has extraordinary engineering application value.In the visual servoing system of manipulator,the acquisition of image information requires the aid of efficient algorithm,which is derived from the rough extraction of camera parameters,cumbersome process of camera calibration,a mass of image noises and errors in the imaging process.Moreover,the visual servoing task is unable to execute successfully due to the constraints of the camera's imaging region,the variation of ambient illumination intensity and the occlusion of images.Furthermore,both the manipulator and the camera are complex nonlinear systems with strong coupling and dynamic uncertainties.From the above discussion,the positioning and tracking control method of the manipulator visual servoing system is developed in this paper by considering the factors such as ambient noise,feature disturbance,field of view constraint and comprehensive energy consumption.This article mainly makes research on manipulator visual servoing positioning and tracking control method with considering the image noise,manipulator visual servoing hybrid positioning control strategy with feature constraints,the manipulator visual servoing double closed-loop tracking control method based on image occlusion/interference,as well as optimal tracking control strategy based on adaptive dynamic programming.The thesis is organized as follows:(1)Since ambient noise and external interference affects the operating accuracy of visual servoing system easily,this article proposes an improved Kalman filtering algorithm to estimate image Jacobian matrix online.Consider the unknown statistical characteristics,the three-segment function is selected to describe the learning statistics,which is regarded as the adaptive factors,and the adaptive robust Kalman filter(ARKF)is employed to dynamically adjust the filtering gain,the noise covariance matrix of the state model is estimated recursively.In the case of non-Gaussian noise,image Jacobian matrix is estimated by maximum correntropy Kalman filter(MCKF),which improves the tracking performance of image feature trajectory.Numerical simulation and comparative analysis are performed for static target and moving target respectively,which further verifies the superiority of the system performance.(2)For the purpose of solving servoing task failure and system divergence caused by the field of view constraints and noise interference in image-based visual servoing control,a hybrid control algorithm combining bidirectional extreme learning machine(B-ELM)and smooth variable structure filter(SVSF)is proposed to approximate the interaction matrix.The mapping function of the image features and interaction matrix is represented by the B-ELM method.The SVSF method is developed to re-estimate the output matrix of the B-ELM method,that enhances the robustness of the control system to noise.Furthermore,the superelliptic boundary is used to smooth the boundary of the view constraint,and a new constraint function is developed based on the hyperbolic tangent function,which can limit the motion speed of image features by dividing regions with different degree of constraint,so as to ensure that the image features are always visible during movement.(3)Focusing on the challenge of visual servoing control subject to feature occlusion and interference,the senarios of image features being occluded or interfered with image features are analyzed.Then,the dual adaptive strong tracking Kalman filter(ASTKF)is presented to adjust the image observation data to effectively estimate the visual state of occlusion/interference and the image Jacobian matrix,which is easy to be apply to the robot platform.Considering the kinematic and dynamic behavior of visual servoing,combining with the uncertainties of the camera and the manipulator model,an outer loop velocity controller of proportional-differential and sliding mode control(PD-SMC)method and an inner loop joint controllerof adaptive sliding mode control(ASMC)method are employed to further enhance the accuracy and robustness of visual tracking.(4)A visual tracking control strategy based on adaptive dynamic programming(ADP)algorithm is employed to solve the subject of energy consumption optimization of manipulator visual servoing systems.A complete model of visual servoing system is built according to the mapping function of feature and torque.The optimal feature error feedback control law is presented by using the critic neural network,which is employed to obtain the ideal tracking control law for fulfilling the visual tracking control.Furthermore,the state observer estimates the overall uncertainties including joint friction,interferences and modeling dynamics in real time.After that,the observed values are introduced into the cost function to improve the overall uncertainties.Then,the optimal image feature error tracking control strategy is derived by combining with the ideal control law.The stability of the manipulator visual servoing system is guaranteed by utilizing Lyapunov stability theorem. |
PDF Full Text Request