Stereo vision-based trajectory following without correspondence information

Abstract

This article addresses the visual servoing of a rigid robotic manipulator under fixed camera configuration. A tool is mounted on the end-effector of a robot which can be controlled by automatic visual feedback. The control goal is to drive the tip of the tool to follow a visually determined three dimensional target trajectory by using a two-camera vision system without assuming any pointwise correspondence information. Based on a novel encoded error, an image-based control law is proposed to achieve precise positioning of the tool in the absence of measurement noise and pointwise correspondence information. With an online falsification algorithm, one can further achieve automatic trajectory following using an approximately calibrated two-camera vision system. The control strategy is successfully validated in simulations on arbitrary line and contour following.