Design of adaptive controller considering disturbances for urban air mobility

Abstract

This thesis proposes an adaptive control system for the multicopter type Urban Air Mobility (UAM) subjected to unknown disturbances such as uncertainties in system parameters and wind, a significant barrier to achieving its mission. First, the effects of uncertainties in a UAM dynamics such as mass, the moment of inertia, and the center of gravity that may degrade the performance are rigorously analyzed. The analysis results show that they appear to be a form of a constant bias or low-frequency disturbance. Based on these observations, the data-driven machine learning-based adaptive law is proposed for disturbance estimation by leveraging the modeling power of Gaussian process regression (GPR). While obtaining a training dataset for learning the GPR model, we utilize the conventional adaptive laws, e.g., time-delay approximation method, and nonlinear disturbance observer, for practicality. The entire control system is integrated by augmenting the disturbance estimation methodology with the baseline control applying the feedback linearization and time-scale separation assumption. Additionally, the control allocation algorithm to coordinate the virtual control efforts and further increase power efficiency is also proposed as our work concerns the over-actuated mechanical system. Finally, the numerical simulations verify the performance of the proposed disturbance rejection control algorithm in the presence of unknown external disturbance. The results present the enhanced performance of the proposed controller when compared with the previous conventional disturbance rejection approaches.