Interval Prediction Models for Data-Driven Design of Aerial Vehicle’s Robust Adaptive Controllers

An important issue in robust adaptive control is the estimation of realistic ranges for the parametric uncertainty to be used in the projection mechanism with the goal of guaranteeing the boundedness of the adaptive parameters in the presence of modeling uncertainty. To facilitate this design it is common practice to assume conservative parameters’ projection bounds and to use high-gain adaptation for reducing tracking errors. However, this approach may not be suitable in practical applications because it might potentially lead to unpredictable transients and peaking phenomena. Therefore, the determination of tighter uncertainty upper bounds is a fundamental problem toward the design of safe adaptive controllers. This study shows that the problem of estimating the parametric uncertainty domain of a linearly parameterized adaptive controller can be advantageously formulated as a linear interval prediction model identification problem. On this basis a practical data-driven method to compute tight upper bounds for the uncertain parameters is proposed through the solution of a convex constrained optimization problem expressed in terms of linear matrix inequalities. The proposed method has been applied to the design of a longitudinal and a lateral directional model reference adaptive control for a fleet of YF-22 research unmanned aerial vehicles based on real flight data.