Robust and Adaptive Control Laws for a mini Quad-Rotor UAV

The scope of this thesis is the implementation of robust and adaptive control theories to guarantee good stability and performance characteristics in formation flight of a multi rotor platform, both for remote piloting and autonomous control. A key point in the evaluation of control laws is the stability dynamic analysis. For this reason a complete mathematical model is implemented, starting from the blade element theory. This model is used for the evaluation of detailed loads in hover conditions. This model can be useful for a modal structural analysis to evaluate the structural frequencies and to customize the data fusion filter with regard to the platform characteristics, to identify the system natural frequencies and to reduce the on board sensor signal noise. Different control laws have been analyzed, from the classical theory, like Proportional Derivative (PD) and Linear Quadratic Regulator (LQR) controllers, to an innovative theory, that is represented by the L1 adaptive controller. The scope is to verify which of these theories is the most suitable for a rotary wing UAV as quadrotor. The validation of controllers is proposed on the experimental model (derived from flight tests) and in a formation flight application. The L1 controller can be implemented to limit the range of the angular velocities and of the four rotor rotational speed. A quadrotor is a platform with fast dynamics on control axes, thus if a sudden maneuver is implemented can cause glitches on the parameters trend and the aircraft could become uncontrollable. Moreover, the structure vibration could increase the platform unstable behavior and the resulting glitches become a variation of the controller variables, thus they are considered as inputs. A key aspect of L1 control theory is the definition of control signals as the output of a low pass-filter, in order to remain in the low-frequency range. This feature permits to avoid high frequency oscillations due to the large adaptation gain; in systems that use electronic devices, like the studied platform, these oscillations significantly increase the current draw and it is undesirable. Moreover, this controller is robust in presence of model uncertainties and unmodeled dynamics. The simple structure and the presence of less oscillations during the implementation demonstrate that this controller can be a better candidate for an autopilot (than the classical control theory). Therefore, a drawback in the L1 controller is the trial and error method to evaluate the low pass filter that is the fundamental component of control law. To provide a systematic method, a mixed deterministic – randomized approach for the control law design (low pass filter) is proposed. The results obtained in the unmatched controller (real platform variables) are thus optimized. In conclusion, these results are validated on a real platform, designed during this research activity and used as test bed for control law algorithms. Quadrotor realization permits to know in detail the platform dynamics and behavior. This is important to better control the system optimizing control law implementation. The implementation of a L1 controller in the on board autopilot can reduce the measurement noise (due to the low pass filter in the control law). This filter can be useful for improving the data acquisition of the accelerometers and for the flight tests combined with the Kalman filter (to have less noisy results).