Visual servoing control of robot manipulator with Jacobian matrix estimation

Abstract

Visual servoing system is a system to control a robot by visual feedback. This paper presents a visual servoing control that drives the end-effector of a real robot manipulator from any arbitrary start position to the desired positions. The control law is obtained using inverse Jacobian matrix. Since there is not access to the model of the robot, artificial neural networks are used to estimate of inverse Jacobian matrix. There are many challenges in practical implementation such as: how to calculate Jacobian matrix, determining the intelligent structure for estimation of Jacobian matrix, recognition coordinate of each joint with image processing and changes in illumination. We proposed appropriate solutions to solve the mentioned challenges. The experimental results in the real robot show that the control system can move the end-effector to target positions from any arbitrary start position with good accuracy.