Adaptive fault-tolerant control of octo-rotor UAV under motor faults in adverse wind conditions
Ibrahim H. Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas · AIAA SCITECH 2023 Forum · 2023
View Video Presentation: https://doi.org/10.2514/6.2023-2535.vid Unmanned aerial vehicles are becoming ubiquitous in domains where safety and robustness against exogenous and endogenous disturbances is critical. In this paper, we discuss fault-tolerant control of an unmanned octo-rotor system under complete failure of a single motor in presence of adverse wind conditions. We combine physics-based and data-driven methods into a hybrid fault-tolerant control scheme that leverages from the history of past flight data to derive locally-optimal control policies. Our control scheme considers a supervisory reinforcement learning (SRL) fault-tolerant controller that defines the reference signals for an onboard hierarchical control architecture commonly implemented in most small aerial vehicles. The goal of the SRL agent is to accommodate motor faults in the presence of flight disturbances, such as wind gusts. We formulate an adaptable control framework by learning from data combined with theoretical robustness guarantees present in classical control approaches. We demonstrate successful results from our approach under different motor failures and varying wind gust conditions in a simulation environment platform based on Gazebo and Python to realize a satisfactory level of realism in the results.