Dynamic inversion based control concept with application to an unmanned aerial vehicle

The paper presents a control strategy for an unmanned aerial vehicle. Based on feedback linearization, a three loop design is performed that finally allows direct commands for speed and flight-path angles. The three control loops correspond to the inversion of rotational, attitude and path dynamics. For the latter, a geometrical approach was employed that gives a more intuitive view on the behavior of the aircraft and the possibility to modify the standard path inversion to improve the tracking accuracy. To exploit the physical capabilities of the design, the maximization of the transmission bandwidth is one of the primary goals. Pseudo-control hedging is utilized in all three loops to account for limitations, saturations and delays in the consecutive inner loops, allowing the aircraft to make use of its full control power including saturated control surfaces. Protection mechanisms were integrated to keep the vehicle in its operational envelope even in rapid and transient maneuvers. Finally, to increase the level of autonomy of the vehicle, the control system was augmented with a neural network to compensate model inversion errors especially in failure situations. The capability of the system is assessed by means of high-fidelity nonlinear simulation including actuation and sensor dynamics or sensor signal processing.