A comparison between gain-scheduling linear quadratic regulator and model predictive control for a manipulator with flexible components

Dynamic performance and energy efficiency can be improved by reducing the inertia of mechanical systems. However, the components of such a system may be flexible and, inevitably, subject to vibrations. Moreover, the relative motion between flexible components leads to time-varying boundary conditions characterizing configuration-dependent dynamics. Since robust controllers may be conservative for this problem, adaptive controllers should be considered since they can cope with performance design requirements. In this work, we present a comparison between a linear quadratic regulator gain-scheduling control and an adaptive model predictive control for a flexible industrial Cartesian manipulator with configuration-dependent dynamics. This study takes the system stability and the use of bounded actuation into account. Numerical results demonstrate that the adaptive model predictive control strategy achieved better performance than the linear quadratic regulator gain scheduled. Also, in applications with tighter actuation bounds, the difference between the performance of both approaches is even higher, reinforcing the model predictive control capability of dealing with limited actuation.