Confidence metrics analysis of a fixed-wing UAV

Abstract

Uninhabited aerial vehicles (UAVs) are becoming popular in the development process of full scale aircrafts and as research platforms. Due to their complexity they provide development and test environments for a wide range of applications. Supporting research projects in safety critical systems, classes, the University of Minnesota Department of Aerospace Engineering and Mechanics have been developing a low-cost UAV research facility. This facility includes models of a family of fixed wing airframes, controllers, a diverse set of guidance algorithms. A flight software is written which implements an autopilot system, including the aforementioned algorithms, and provides datalogging. The software package is equipped with tools to evaluate flight test results. The model of any plant is never 100\% accurate. There are always differences between the real system and the dynamical model of it. Uncertainties can be introduced into the model, which are trying to capture uncertainty in model parameters and unmodeled dynamics. Even though the aircraft model in the package is fairly accurate, it is interesting to investigate 'how good' the model is, i.e. how robust the model in the closed loop is against uncertainties. Earlier work in this project mainly focused on plant modeling and controller design \cite{dorobantu_20133, paw_2009}. Extensive controller analysis, however, has not performed yet, what motivates the work behind this thesis.