Optical-Flow-Based Visual Servoing for Robotic Moving Control Using Closed-Loop Joints

This paper studied an optical-flow-based visual servoing method for eye-in-hand robotics task manipulation without system calibration. Firstly, considering the object feature points are easy to be lost with the traditional Lucas-Kanade(LK) optical flow method in the process of camera fast moving. We propose an improved pyramid optical flow tracking algorithm which iterates the vision image by layered, this algorithm obtains a hierarchy image by building a Gaussian pyramid, which can reduce the image motion speed and improve the tracking robust performance. In this work, we need to use the information obtained in the optical flow tracking process for visual servo control to drive the robot to complete the task manipulation, thus we use joints closed-loop controller and joint-image Jacobian matrix to construct a new visual servoing system for robotic motion control based on improved optical flow method. Finally, we have completed the visual servoing task manipulation on an eye-in-hand uncalibration robotics with six degrees of freedom (six-DOF). The experimental results show the effectiveness and robustness of the proposed scheme.

1 Introduction

In recent years, robotics had been widely used in industrial tasks in virtue of their high efficiency and accuracy, especially in the assembly industry ^[1]. However, most of the robots are controlled by teaching device in the market, which limits their working path and makes robot lost their flexibility. While with the development of vision sensor, the robot equipped vision can accomplish more dangerous and accurate positioning tasks, such as underwater exploration ^[2]. Therefore, the purpose of this paper is to use the image tracking algorithm and visual servoing closed-loop controller to control an uncalibrated six-DOF robot to complete the positioning and grasping tasks in real world.