Adaptive robust attitude controller design for a quadrotor platform

Yılmaz, Emre
This thesis includes attitude controller design ideas for a quadrotor platform which can be regarded as an exceptionally agile flying robot with highly nonlinear and unstable features in flight dynamics. These platforms pose severe problems in characterizing the dynamics especially when performing high-speed manoeuvres. These facts cause the quad-rotor not to lose its popularity as a compelling tool among avid researchers who endeavour to realize various controller ideas. The procedure in this thesis is initiated with the construction of the system model and the verification of this phase relying on the characteristics of the test bed. With the aid of sensors on the off-the-shelf platform, the controllers are designed to enact tracking of the reference commands that contain the desired trajectories and attitudes. The controller methods highlighted in this research are non-linear dynamic inversion, model reference adaptive control and integral back-stepping technique. The trade-off between performance and robustness is investigated as well. The responses of the system to the impacts of the existence of uncertain parameters, unmatched uncertainties or disturbances are exceptional means to judge how robust the controller is. An overview of the cases with parametric uncertainty and the existence of noise, therefore, find its place as a section within this work. This sketch grades the controller options while putting forward the advantage of adaptation. The simulation results show that all controllers operate Exceptionally in noiseless and noisy scenarios. Under the cases where high-level parameter uncertainties or unknown disturbances exist, however, the functionality of base controller and Integral Back-stepping can be claimed no more. In coherence with its purpose of integration into the controller, only adaptive algorithm survives these situations. The analysis including the computation of 2-norms and maximum absolute value of error vectors in the commanded state axis brings about supportive results for the deductions about the adaptation. Besides, by employing correction approaches, the advancement of the controller in terms of robustness is examined where dead zone implementation, e- and sigma-modifications are exploited. Among these modifications, e-mod and dead zone satisfy specified criteria for convergence and robustness while sigma-mod is determined as useless. In the procedure, studies about simulations with various step size values, various fixed-step ODE methods, different levels of unknown disturbances and uncertain parameters are also conducted in order to see the sensitivity of the adaptation against these criteria. The reliability of the methodologies should be justified through experiments and the analogy between experiment and theory is provided. The motivation behind this research is to produce persistent attitude controllers to lay the first stone for more complex algorithm structures such as autonomous flight phases, obstacle avoidance and way-point targeting.