Research On Uncalibrated Hand-eye Coordination Motion Tracking Control Of Industrial Robot

In the robot hand-eye coordination,uncalibrated robots overcome the defects and difficulties in practical application based on calibration methods,thus becoming the research focus of researchers.In this thesis,a six degree-of-freedom robot is taken as the research object,aiming at the motion tracking control problem in the robot hand-eye coordination,the UPF particle filter method of state estimation and the model predictive control method are respectively used for research,and the main research contents are as follows:(1)Firstly,in order to study the motion tracking control problem in the robot hand-eye coordination system,the kinematics model of the six degree-of-freedom industrial robot UR5 is established,the forward kinematics and reverse kinematics of the robot are analyzed and solved,and the forward and reverse kinematics are simulated experimentally.(2)Secondly,aiming at the robot uncalibrated hand-eye coordination,the unscented particle filter is used to estimate the Jacobian matrix online,and the definition of the Jacobian matrix is given,and the online estimation of the Jacobian matrix is transformed into a Bayesian filter framework.Then,an unscented particle filter is proposed to estimate Jacobian matrix.The unscented Kalman filter equation is used to propagate and update each particle.This method can make full use of the observation results of the current image feature quantity,and then produce an achievable estimation result.Simulation results on a six degree-of-freedom robot platform show that the method improves the tracking effect and reduces the tracking error in two-dimensional motion tracking and three-dimensional motion tracking tasks compared with commonly used PF and UKF.(3)Finally,aiming at the limitation that UPF online estimation of the total Jacobian matrix can only be applied to the eye-in-hand,a perspective projection model which can unify the two camera configurations is used to study the uncalibrated hand-eye coordination when the internal and external parameters of the camera are uncalibrated and the characteristics are unknown in the three-dimensional space position parameters.Based on the model predictive control method,constraints of the robot system,such as visibility constraints,are explicitly considered.Most IBVS controllers use traditional image Jacobian matrix.The IBVS scheme proposed in this thesis is implemented by using depth independent interaction matrix.Unknown parameters can appear linearly in the prediction model.Identification algorithm can effectively estimate unknown parameters.In addition,the model predictive control determines the optimal control input and updates the estimation parameters together with the predictive model.This method can handle system constraints,unknown camera parameters and depth parameters simultaneously.Moreover,the motion tracking task can achieve better performance.Finally,two-dimensional motion and three-dimensional motion simulations are carried out with a 6-DOF robot in the eye-in-hand and fixed-eye configurations,respectively.The results verify that the new IBVS method can be used in both eye-in-hand and fixed-eye camera configurations,and can improve the motion tracking control accuracy and reduce the motion tracking error of motion tracking control compared with UPF state estimation method.