Cartesian base predictive control of robotic manipulators

Abstract

A multi-input multi-output (MIMO) controlled auto regressive integrated moving average (CARIMA) model which has the input vector consisting of the joint torques and the output vector consisting of the velocities of the gripper expressed in the world coordinate system, is proposed for the motion of the end-effector of a robotic manipulator system. Since the manipulator is a time-varying system, because of changing of the load and the friction between the joints the proposed MIMO model parameters are updated recursively online by the least squares (RLS) algorithm. A self-tuning type predictive controller is then designed by minimizing a quadratic performance criterion which includes the position error of the gripper and the energy associated with the torque vector, to control the position of the end-effector. The performance of the predictive controller is tested under varying operating conditions by digital simulation studies.<>