Uncalibrated Visual Servoing Using More Precise Model

Abstract

A visual servoing method using a more precise model is presented in this paper. It uses a secant to approximate the second order term in the Hessian of the Newton model. This is different from the popular so-called quasi-Newton uncalibrated visual servoing method, which neglects the second order term directly. Its performance is superior to the so-called quasi-Newton uncalibrated visual servoing method, especially in large residual cases. To guarantee the global convergence of this method, a trust region method is used. Besides, a recursive least squares algorithm is employed to estimate the coupled image Jacobian, so it is not necessary to know the parameters of the camera and the robot. More than that, an approach to improving the control precision of the end-effector in the workspace is also proposed. In the end, a three-degree-of-freedom robot with two fixed cameras system is simulated to validate the method. The simulation results demonstrate the effectiveness of the method.