High performance 6 DOF visual servoing using Generalized Predictive Control

This paper presents a new approach to high speed visual servoing in the case of a 6 DOF industrial manipulator that takes into account the dynamics of the manipulator in the synthesis of the visual controller. The manipulator with its actuators (DC motors), their current feedback loops and their velocity control loops, is modelled as a "virtual Cartesian motion device". A linearized model of the dynamics is identified around working configurations of the manipulator. Then, based on this higher order model of the dynamics of the manipulator, 6 Generalized Predictive Controllers (GPC) are implemented on line. Simulations and experimental results show a drastic improvement in performance for a 6 DOF industrial manipulator in an eye-in-hand configuration with a high speed visual servo-loop (120 Hz) compared to standard approaches neglecting these dynamics.