A New Learning Controller For Mechanical Manipulator

Abstract

A new learning controller for motion control of mechanical manipulators undergoing periodic tasks is developed. This controller does not require exact knowledge of the manipulator dynamical structure or its parameters, and is computationally efficient. In addition, no actual joint acceleration is needed in the control law, and the full nonlinear dynamics of the manipulator is fully accounted for in the convergence proof. Simulation results of this algorithm applied to a realistic Scara type manipulator [8], which includes friction, pay-load inertia variations, actuator/sensor noise, and unmodeled dynamics is also presented.