Visual servoing in robot motion control

The paper describes some basic principles and derived methods for the visual servo modelling and control of rigid robot manipulators. Guidance vision is introduced as an advanced motion control method, which provides flexibility when integrating industrial robots in computer-controlled manufacturing structures. Two important system architectures are analyzed: dynamic look-and-move systems, and direct visual servoing systems for stationary and arm-mounted cameras. The generic robotic task modelled in the paper is the visual tracking of objects for robot grasping. Motion control tasks are modelled by Virtual Kinematic Error functions and the composition of relative poses extracted from image features. The paper reports experiments carried out on two SCARA-type robots and the analysis of robot-object positioning errors, specific to object grasping and manipulating tasks.