Hybrid Jacobian control for uncalibrated robot visual servoing

Abstract

This paper focuses on the visual servo control of an uncalibrated robotic arm with an eye-in-hand camera. Without a prior knowledge of the kinematics of the robotic arm or camera calibration, the proposed hybrid Jacobian controller can track a moving object using visual feedback and joint-space velocity feedback. The proposed hybrid control method is a combination of the uncalibrated visual servoing and approximate Jacobian feedback control. First, the Jacobian matrix from joint-space to image-space is estimated by recursive least-squares (RLS) algorithm, and then the approximate Jacobian feedback controller is designed by using visual feedback and joint-space velocity feedback. The performances of the proposed control methods are illustrated by computer simulations.