Model Free Adaptive Control under Gradient Method for Estimation of Energy and Pseudo-Partial-Derivative Factors
Edwin Calla Durandal1; Guilherme Americo Rosa1; Rejane de Barros Araujo1; Antonio Augusto Rodrigues Coelho1
1 Federal University of Santa Catarina
Resumo
The Model Free Learning Adaptive Control (MFLAC) is based on the pseudo-partial-derivative (PPD) calculated from the input and output signals of the system to be controlled and also using a fixed penalty factor that weights the control energy to provide a good behavior to the feedback process. This paper presents a new adaptation algorithm to adjust not only the PPD but also the energy factor (λ), both based on the Classical Gradient method (CG). With this proposal, an alternative control design is obtained for the on-line tuning of both parameters, ensuring good stability and robustness for the closed-loop system. Two numerical simulations on SISO discrete-time non-linear plants demonstrate the efficiency and superiority of the proposed adaptive controller when using an estimation mechanism to adjust these parameters over the standard MFLAC. Performance indices are used to show the improvement of the behavior of the proposal control algorithm.
Palavras-chave: Model free adaptive control; penalty/energy factor; pseudo-partial-derivative; gradient method; non-linear system; autotuning