Research On Target Grasping Based On Hand-Eye Coordination For Underwater Vehicles

The autonomous operation of the Underwater Vehicle Manipulator System(UVMS)is playing an increasingly important role in sea organism capture tasks.Hand-eye coordinationbased control is one of the key technologies for UVMSs to achieve autonomous operations,and researches on hand-eye coordination-based control are important to enhance the level of autonomous and intelligent operations of UVMSs.Compared with the general land-based manipulators,the UVMS often has high degrees of freedom due to its multi-body mechanism,hence,the coupling effects between the manipulator and the vehicle are complex;besides,the UVMS kinematic parameters and camera parameters are difficult to accurately achieve;moreover,the UVMS can be easily affected by external environmental influence in the underwater environment.Therefore,traditional visual servoing methods for land-based manipulators may not suitable for the UVMS.The coordination between the UVMS and the underwater camera,i.e.,the coordination between the "hand" and the "eye" of the system,needs to be fully considered to achieve autonomous operations based on visual feedback information.This paper analyses some of the difficulties in hand-eye coordination-based target grasping by the UVMS and conducts a study on this issue,which are mainly as follows.The UVMS hand-eye kinematic model is established by analyzing both the generalized velocity relationship and the generalized displacement relationship between the end-effector and the UVMS,the perspective camera model,and the relationship between the velocity of the feature point on the image plane and the camera velocity.The modeling and analysis of the vehicle dynamics model.the coupling effects between the manipulator and the vehicle,the UVMS dynamics model,and the end-effector dynamics model are performed,laying the foundation for the subsequent development of the coordinated control algorithms of the UVMS.The UVMS hand-eye coordination-based grasping methods by using "eye in hand" and"eye to hand" cameras are investigated,respectively.In the case of the "eye to hand" camera,the novel motion controller in the horizontal plane of the vehicle is constructed by using the state of a single feature point on the image plane,which enables autonomous grasping during UVMS hovering.In the case of the "eye in hand" camera,a novel coordinated motion controller based on the dynamic model of the UVMS coupling effects is constructed,and the stability of the system is proven.Hence,an image based visual servoing method combined with the novel coordinated motion controller is utilized to realize autonomous grasping during UVMS hovering.Experiments based on the UVMS in pool environment following the proposed methods demonstrated the effectiveness of mentioned hand-eye coordination grasping methods.The UVMS hand-eye coordination-based grasping for the stationary target by using a binocular camera is investigated.First,UVMS coordinated controllers are designed in both configuration space and task space based on the Sigmoid function,respectively.Second,a modified nonlinear disturbance observer is given,and the stability of the coordinated motion controllers combining with the disturbance observer is proved based on Lyapunov theory.Finally,a finite-times switching strategy combining with two types of coordinated controllers is proposed to reduce the error of the end effector to achieve accurate target grasping of the UVMS.Simulations are carried out by using a 10-degrees-of-freedom UVMS in three cases,and results show that the proposed method can deal with external disturbances to reduce the end-effector residual errors and thus achieve accurate target grasping.The UVMS hand-eye coordination-based grasping of the stationary target by using an "eye in hand" camera without calibrations is investigated.First,a hand-eye kinematic decomposition algorithm is proposed,which can decouple unknown parameters of the system into constant vectors,and overcome the assumption in previous works that the kinematic parameters of the manipulator need to be obtained in advance.Then,the decomposition algorithm is utilized to design adaptive laws to estimate unknown vectors,and a generalized reference velocity update method that considers the restoring moments optimization of the UVMS is designed in configuration space.Finally,a novel hand-eye coordinated controller is designed based on a high-order disturbance observer to resist external disturbances and internal dynamic uncertainties,and the system stability is proved by using Lyapunov theory.Simulations are conducted under the conditions of unknown the UVMS kinematics parameters,camera parameters,the target position information while suffering external disturbances.Case study 1 shows that the proposed method can optimize the restoring moments of the UVMS while effectively tracking the desired trajectory on the image plane,and maintain robust against external disturbances.Multiple feature points situation is demonstrated in case study 2,the endeffector can accurately reach the desired position and orientation in the Cartesian space,which illustrates the proposed method is effective in underwater autonomous operations.The UVMS hand-eye coordination-based grasping of the moving target by using multi cameras is investigated.First,an improved adaptive Broyden's class algorithm is proposed to estimate the residual matrix in the monocular camera-based visual servoing process,the convergence analysis of the proposed algorithm is also given.Afterwards,considering the camera field of view limit and UVMS displacement limit,a recursive optimization of the generalized displacement of the UVMS is proposed.Then,a systematic Jacobian matrix fusion algorithm is proposed to decrease the track unevenness caused by the trial motion based on multiple cameras.Finally,in order to realize long range grasping,a hybrid hand-eye coordination strategy is proposed based on the multi cameras of the UVMS,the switching criteria are also presented to prevent high-frequency switches.Simulations of the proposed uncalibrated visual servoing scheme based on a monocular are carried out,compared with several classical algorithms,the proposed adaptive Broyden's class algorithm demonstrates better performances in case study 1.In case study 2,simulations are also carried out to verify the robustness and effectiveness of the proposed hybrid visual servoing scheme based on multi cameras,multiple switches are introduced according to the routine of the algorithm.After that,simulations of the hybrid hand-eye coordination strategy combining with a dynamic based coordinated motion controller to testify its applicability and robustness.