CAMERA CONFIGURATIONS OF A VISUAL SERVOING SETUP, FOR A 2 DOF PLANAR ROBOT Paulo J. Sequeira Gonçalves 1 J. R. Caldas Pinto 2 1 Escola Superior de Tecnologia de Castelo Branco, Av. Empresário, 6000-767 Castelo Branco, Portugal 2 Instituto Superior Técnico, Av. Rovisco Pais, 1096 Lisboa, Portugal Abstract: This paper presents a study of three different camera configurations, for a 2 dof planar robot, comparing their behavior based on the singularities of the jacobian used in the control law. The camera configurations used were: eye-to-hand; eye-in-hand with the camera looking respectively in front and down. Two image features, coordinates of a point, from the target were used. Plots representing jacobian singularities, within the joint limits, are presented for a known target position. Some conclusions are also drawn when the target position is unknown. Simulation results for 2D Visual Servoing are presented, to show the behavior of the servoing in the three camera configurations. Copyright © 2003 IFAC Keywords: Visual Servoing; Robot Vision; Robotics Manipulators; Computer Vision; Robot Control. 1. INTRODUCTION In this paper we control, based on 2D Visual Servoing and under different camera configurations, a 2 dof planar robotic manipulator. The robot in use was constructed at Instituto Superior Técnico (Martins and Sá da Costa, 1999; Baptista, et al, 2001), Mechanical Engineering Department, Robotics and Automation Laboratory. Planar robots are widely spread in the industry, two of their most important characteristics being speed and precision. This type of robotic structure can easily evolve to a SCARA robot, used in vertical assembly tasks, palletizing, etc. The first objective of this work is to study the planar robotic manipulator under 2D Visual Servoing, for three different camera configurations: eye-to-hand; eye-in-hand with the camera looking in front and down. The second objective is to find the singularities of the jacobian, relating the image features velocity to the joint velocity, in order to avoid them in path planning schemes. The robot should move from an initial to a final position, with the control taking place in the image features space. The control law and the control structure used to accomplish these objectives are presented in section 2. The planar robot manipulator is presented in section 3. The three camera configurations and the singularities of the jacobian are presented in sections 4 and 5, respectively. The simulation results are presented in section 6. Some conclusions of this work are exposed in section 7. 2. VISUAL SERVOING Machine Vision and Robotics can be used together to control a robot manipulator. This type of control, defined as Visual Servoing, uses visual information from the work environment to control a robot manipulator performing a task. The visual information can be obtained by two ways: using direct information from the image (2D visual servoing); or using 3D information of the object from the image(s) (3D visual servoing). The second case needs an on-line processing for pose estimation. A good explanation of the differences is done in (Hutchinson, et al., 1996). Combining the previous approaches (Malis, et al, 1999) have proposed the 2½D Visual Servoing, nowadays also called Hybrid Visual Servoing. In this paper only 2D Visual Servoing is used.