Non-linear Model Predictive Control for Longitudinal and Lateral Guidance of a Small Fixed-Wing UAV in Precision Deep Stall Landing

This article presents the use of constrained non-linear model predictive control (NMPC) for high-precision deep stall landing of a fixed-wing unmanned aerial vehicle (UAV), through simulations in six degrees of freedom. The control algorithm takes the UAV horizontally in the correct direction of the landing target until the longitudinal line of sight angle from the UAV to the landing target has reached a given limit. Then it descends the UAV directly towards the landing target, keeping a steep flight path and enforcing a deep stall. The performance of the landing scheme has been evaluated with different path angles and wind velocities, where wind measurements from the last time step are assumed known by the NMPC. This showed acceptable landing performance in winds up to 7 m/s, and indicated that steeper flight paths are more susceptible to wind disturbances. To challenge the assumption of fast and accurate wind estimates, a simulation where the gust wind was assumed unknown by the NMPC was performed. This reduces the performance of the landing controller, but the majority of the feasible paths successfully land in a deep stall within 2 meters of the target.