Experiments and simulation of model-based controlled robots

Abstract

This paper presents an overview of a part of the work done in the Robotics and CAD Laboratory on the computational and experimental aspects of feedback control of mechanical manipulators. We first present experimentalresults which show that a feed forward, model-based controller performs significantly better than a proportional plus derivative (PD) controller. The experiments were performed at the Bhabha Atomic Research Centre (BARC),Bombay, on a complex five-degree-of-freedom manipulator, containing a four-bar mechanism for motion transmission,large speed reductions with accompanying friction and backlash at joints, and driven by AC servo motors.We next present simulation results which show that a simple two-degree-of-freedom manipulator undergoing repetitive motion, under model-based or PD control, can exhibit chaotic motions for a particular range of feedback gains and for large mismatch between the model and the actual parameters.