Data-driven direct automatic tuning scheme for fixed-structure digital controllers of hybrid systems

Abstract

In this study, a novel data-driven direct automatic tuning scheme is proposed for the general form of fixed-structure digital controllers of hybrid systems. The controller is directly auto-tuned while it interacts with the real plant; no modelling or plant identification procedure is involved. The auto-tuning problem is formulated as a controller parameter optimisation problem, which is solved using a quasi-Newton method. A new method for calculating the gradient of the objective function from closed-loop experiment data is proposed to reduce the sensitivity to the initial values of the parameters. Domain knowledge of controller design based on pole placement and internal model principle is introduced as a priori information to further improve the performance of the auto-tuning. Numerical simulations are conducted to demonstrate the effectiveness of the proposed scheme and performances of different objective functions are evaluated. The performance of the proposed scheme is compared with the standard relay feedback and iterative feedback tuning methods. Reliable controllers can be obtained automatically by using the proposed scheme; the only remaining task of the control engineer is to estimate the order of the plant by counting the number of independent energy storage elements.