Nonlinear Dynamic Inversion and Neural Networks for a Tilt Tri-Rotor UAV

A small scale tri-rotor test bed with tilting propellers has been built to test flight control laws in view of the construction of a larger tilt rotor UAV. As a first step to achieve autonomous flight capabilities, a nonlinear dynamic inversion based flight controller is developed. This controller is designed on the basis of a time-scale principle with two levels. A lower level, fast control action, designed to achieve attitude control and stability goals, is driven by a higher level trajectory tracking control law. To achieve robust stability and performance in the presence of parametric variations and modelling uncertainties, an adaptive flight control law correction based on neural networks is investigated. A RBF neural network is implemented to mitigate the effects of imprecise inverse dynamics. The overall proposed flight controller performance are tested via numerical simulations on the mathematical model of the small scale tri-rotor. Preliminary results on the full tilt rotor are also shown.