Experimental Validation of a Predictive Control Co-Design Algorithm

Abstract

Control co-design (CCD) represents a promising solution for coordinating the physical design and control of dynamic engineering systems as technological demands become more stringent. Predictive control co-design (pCCD), recently introduced to the CCD literature, optimizes combinations of feedforward and feedback static gain sets at the system design stage to combine the robustness and preview control afforded using state-of-the-art control methods, like model predictive control (MPC), in CCD with the computational efficiency of open-loop CCD methods that solve CCD problems with a single optimization level. This work contributes the first experimental validation of pCCD to the literature. First, pCCD is performed offline on a spring-mass-damper system. The co-designed system’s optimal response is then experimentally validated online. Results are compared to an analogous system co-designed with an open-loop CCD method. The experimental system co-designed using pCCD yielded a sum squared error with respect to a desired reference signal 40 times smaller than the system co-designed using open-loop CCD. The results indicate that pCCD yields co-designed systems with superior online robustness in comparison to open-loop CCD methods. Moreover, systems co-designed using pCCD are more robust to both modeling error and unexpected disturbance inputs or changes in desired reference signals encountered online.