Data-driven set-point tuning of model-free adaptive control

Model-free adaptive control (MFAC) is an effective data-driven control method to deal with nonlinear and nonaffine systems. In this article, a data-driven set-point tuning (DDST) approach is proposed for MFAC to enhance its control performance. The proposed data-driven set-point tuning based MFAC (DDST-MFAC) system consists of two control loops. The inner control loop takes the MFAC as the feedback controller where a virtual reference error signal is adopted in the control input. The DDST in the outer loop is derived from an ideal nonlinear set-point tuning (NST) law, which exists in theory, to meet the control target. To realize the theoretically existing NST, a dynamic linearization (DL) technique is introduced. Virtually, the ideal NST law is independent of any controlled system, regardless linear or nonlinear, having an exact model or not. Since the nonlinear system considered in this work does not have any model information available to the designers, the parameter estimation law of the DDST is designed by using the DL method to transfer the original nonlinear system into a linear data model (LDM) with a projection algorithm to estimate the unknown parameters in the LDM. The convergence of tracking error is proved for a regulation scenario. Simulation study is provided to verify the theoretical results.