Precision Deep-Stall Landing of Fixed-Wing UAVs using Nonlinear Model Predictive Control

Abstract

To be able to recover a fixed-wing unmanned aerial vehicle (UAV) on a small space like a boat deck or a glade in the forest, a steep and precise descent is needed. One way to reduce the speed of the UAV during landing is by performing a deep-stall landing manoeuvre, where the lift of the UAV is decreased until it is unable to keep the UAV level, at the same time as the drag is increased to minimize the speed of the UAV. However, this manoeuvre is highly nonlinear and non-trivial to perform with high precision. To solve this, an on-line nonlinear model predictive controller (NMPC) is implemented to guide the UAV in the landing phase, receiving inputs from the autopilot and guiding the UAV using pitch and throttle references. The UAV is guided along a custom path to a predefined deep-stall landing start point and performs a guided deep-stall. The simulation results show that the NMPC guides the UAV in a deep-stall landing with good precision and low speed, and that the results depend on a correct prediction model for the controller.