Neuroadaptive Observer-Based Fault-Diagnosis and Fault-Tolerant Control for Quadrotor UAV

This paper highlights the design of an adaptive observer based on radial basis function neural network (RBFNN) for the diagnosis of actuator faults in a quadrotor UAV (Unmanned Aerial Vehicle), and the design of a fault-tolerant control system. First, the quadrotor UAV model considering the faults and unknown disturbances is presented. Next, the adaptive observer design is implemented using a reduced order model of the system, in order to simplify the structure of the neural network and consequently reduce the computational complexity and facilitate the adjustment of its parameters. The fault-tolerant control system is then exposed through the combination of the controller and the faults estimated by the adaptive observer. Simulations are performed in order to demonstrate the effectiveness of the proposed system with respect to the speed and accuracy of faults estimation with nonlinear behaviors and simultaneous occurrences in the actuators, and the attenuation of the effects of faults on the quadrotor UAV dynamics.