Fault Diagnosis And Fault Tolerant Control Of Quadrotor Uavs

Quadrotors represent a special class of Unmanned Aerial Vehicles (UAVs) and have attracted significant attention in recent years due to their potential in various military and civilian applications. However, due to their intrinsic fabrication process and component damage, quadrotors are prone to various type of faults. This dissertation presents the design, analysis, and experimental implementation of fault diagnosis, fault-tolerant control, and controller verification methods to achieve safety assurance and trusted autonomy of quadrotor UAVs. First, the issue of sensor faults is investigated under two different scenarios: (1) the case when all Euler angles are available for measurement; (2) the case when roll and pitch angles are not measurable and need to be estimated. Nonlinear adaptive estimators are designed to provide possible simultaneous accelerometer and gyroscope fault detection, isolation, and bias estimation. Next, the issue of fault-tolerant control of quadrotor UAV in the presence of actuator faults is considered. First, the design of an integrated fault diagnosis and accommodation scheme is investigated. Nonlinear adaptive thresholds are designed to improve the robustness of the fault detection and isolation algorithm. The fault diagnostic information is used for accommodating the effect of the faults. Second, a nonlinear adaptive fault-tolerant altitude and attitude controller is developed to automatically stabilize and recover tracking performance of the quadrotor, even in the presence of possible multiple simultaneous actuator faults. Compared with the first approach, the adaptive control framework is able to automatically accommodate the effects of actuator faults without the need of a fault diagnosis mechanism. Finally, a run-time assurance architecture is investigated for the verification and validation of the adaptive fault-tolerant altitude attitude controller. The algorithms are presented with a rigorous framework aimed at characterizing their performance properties. The above algorithms are implemented and evaluated using a real-time indoor quadrotor test environment. Experimental flight test results are shown to illustrate the effectiveness of the proposed methods.