Enhanced generalized predictive control with disturbance compensation

Abstract

This paper presents an improved generalized predictive control (GPC) scheme integrated with a disturbance compensation scheme that combines iterative learning control (ILC) and real-time feedback control (RFC). A least mean square error (LMSE) estimator has been used to estimate the output error caused by repeatable disturbances. The use of this estimated error information in the ILC component aims to reduce the effect of real-time disturbances in the learning process. On the other hand, the inclusion of the current cycle error information, handled by the RFC component, allows the controller to make more immediate corrections with respect to disturbances occurring during the on-going operation. The proposed GPC-ILC-RFC-LMSE method is simulated on a two-link planar robotic manipulator that is to track a circular trajectory repeatedly. A discrete-time model of the robotic manipulator is used to predict the system output over a prediction horizon such that optimal control inputs that minimize the angular position and velocity trajectory errors can be determined. The proposed GPC-ILC-RFC-LMSE scheme succeeds to reduce the trajectory tracking errors significantly where the average MSE values is merely 40% of that of the GPC-ILC counterpart. In addition, the proposed controller is more robust if compared to the existing GPC learning methods where smoother control input profiles has been achieved.