Model-Based Robust Control of an Ultralight Fixed-Wing Tailsitter

Abstract

Development of fixed-wing VTOL UAVs is an active area of research, enabling increased range and top-speed far surpassing conventional quad-rotors, while preserving the landing flexibility. A subset of this research is the fixed-wing tailsitter, which achieves vertical landing by pitching the entire airframe up into hover. This thesis investigates model-based robust control of an ultralight 250 gram tailsitter UAV which aims to autonomously land vertically in real-world conditions. A 6-DOF model adapted for the unique challenges of tailsitters at hover was developed, extracting aerodynamic coefficients through CFD simulation. By incorporating propwash dynamics, the model ensures control surfaces maintain authority as airspeed approaches zero, enabling continuous simulation from cruise to hover. To account for the significant nonlinearities during transition, an LPV model was developed, scheduled over a grid of airspeed and angle of attack to cover the flight envelope. From this, a MIMO H∞ controller was developed to stabilize the vehicle’s coupled dynamics. Simulation results demonstrate successful trajectory tracking through transition and descent, showcasing that the LPV captures the nonlinear dynamics. Furthermore, wind gust simulations utilizing the Dryden wind model were used to evaluate the operational limits of the aircraft. These tests revealed that the gain-scheduled LQR baseline demonstrated higher overall performance and superior disturbance rejection. Ultimately, the results indicate the aircraft can only handle low to moderate wind conditions, which is attributed to its low inertia and control authority.