Modelling and controlling of an autonomous tail-sitter vertical take-off and landing (VTOL) unmanned aerial vehicles (UAVs)

Abstract

Tail-sitter vertical take-off and landing (VTOL) unmanned aerial vehicle (UAV) combines the advantage of simple mechanism, easy takeoff and landing, high-speed cruises and long flight endurance. These advantages have attracted the attention of researchers, but it is still required to overcome the difficulties of flight control for utilizing this kind of vehicle. In this study, model predictive controllers (MPCs) were proposed to control the position of a quadrotor tail-sitter VTOL UAV. The vehicle was manufactured and it's dynamic and kinematic was described in the inertial frame of north-east-down (NED). The propulsion system was modeled by setting up a thrust experiment. The aerodynamic effect of the wing was modelled by applying the momentum theorem and the component breakdown method. Then a successive linearization MPC (SLMPC) controller was designed for hovering control based on a plant model, an estimated disturbance model and an unmeasured disturbance model, followed by the determination of the objective function and the constraints on variables. The SLMPC controller was tested and tuned under a software-in-loop (SIL) condition until a stable and satisfactory performance. Then it was installed onto the vehicle and indoor flight experiments of disturbance rejection were conducted in the Vicon motion capture environment. The result has shown that the proposed SLMPC control method can perform a precise and stable position holding under non-uniform windy conditions compared to the traditional linear MPC (LMPC) controller. A system identification method is taken to model the vehicle under the cruise stage. Grey box models were derived from the dynamic and kinematic equations and the outdoor flight experiment was designed. An identification section and a validation section were selected from the collected flight data, followed by feeding to a low-pass filter. A least square regression method is taken to fit the grey box model to the data of the identification section and the problem is solved by using a trust-region algorithm. By obtaining and validating the models in the longitudinal and lateral direction, MPC controllers have been set up and tested in the SIL environment. A controller switching mechanism is then developed to complete the large envelope control of the forward and backward transition. During the controller switch, a warm-up mechanism can help to suppress the unwanted control chattering. They have been tested the SIL condition and the improvement in transition performance is presented in the result.