Dynamic Linearization Residual-Assisted Model-Free Adaptive Control With Modified Criterion Function Based on Extended FFDL Data Model

In practical applications, control accuracy and dynamics are two of the critical metrics for evaluating the performance of control systems. To better address these requirements, a data-driven control method known as dynamic linearization (DL) residual-assisted model-free adaptive control is proposed in this article. By combining the impacts of external disturbances and internal uncertainties through the DL residual, an extended full-form DL (FFDL) data model is introduced, based on which, a DL residual-assisted compensator is proposed correspondingly to enhance the accuracy. For the improvement of dynamics, a differential term of plant output is incorporated into the modified criterion function for controller output. Additionally, online pseudo gradient estimation is conducted by using the recursive extended least squares method, which ensures more rapid convergence to the actual value. The stability of parameter estimation and the bounded-input-bounded-output (BIBO) stability of the control system are demonstrated. Finally, numerical simulations and experimental validations confirm the competitiveness of the proposed scheme.