Cинтез Робастного Під-Регулятора Для Невизначеної Нелінійної Моделі Бпла З Використанням Штучної Моделі Посилення І Тимчасової Затримки

Abstract

Estimations of the flight vehicles aerodynamic coefficients through the theoretical, numerical, wind tunnel and flight-test methods have always errors and uncertainties. Nonlinear dynamics of the unmanned aerial vehicle due to speed variations, as well as variability and uncertainty of the aerodynamic coefficients can be considered as an uncertain model. A robust proportional-integral-derivative controller is designed based on the nonlinear optimization in the time domain for the uncertain nonlinear dynamical model of the unmanned aerial vehicle. Artificial gain and time-delay models are added to achieve required stability margins as system robustness during the robust proportional-integral-derivative control design. The aerodynamic coefficients are divided into two groups on the basis of the output vector's sensitivity to the aerodynamic coefficients. Nonlinear optimization of the criterion is performed in two steps for the robust proportional-integral-derivative controller design. In the first step of the design, nominal values are used for coefficients with low-sensitivity, and upper and lower limits are used only for high sensitive aerodynamic coefficients. In the second step, the upper and lower limits are applied all of the aerodynamic coefficients to evaluate and re-adjust the robust proportional-integral-derivative controller parameters. The robust controller is designed for the uncertain nonlinear roll and lateral dynamic model of the Skywalker X8 flying wing to show the effectiveness of the proposed method to guarantee the stability margins. In the given example during the design of the robust controller, with adding of the artificial gain and time-delay model in the control loop the controller parameters are changed. This regulator increases the phase stability margin by about 10 degrees and the gain stability margin by about two for the family of the equivalent uncertain linear models.