Fault-Tolerant Control for Unmanned Aerial Vehicles via an Adaptive Neural Network Observer

To tackle the problem of multiple types of faults for quadcopter unmanned aerial vehicles (UAV), the paper proposes a fault estimation observer and fault-tolerant controller based on an adaptive radial basis function (RBF) neural network. This observer is devoted to addressing a nonlinear attitude control system of a quadcopter UAV that exhibits dual faults in both actuators and sensors. The paper decomposes the original system into two subsystems: one exclusively accommodating actuator faults and the other solely addressing sensor faults. The paper then designs an adaptive RBF neural network observer for the subsystem with actuator faults only. This approach extends the sensor fault to the form of system state, thereby constructing an adaptive observer to accomplish fault estimation. An fault-tolerant controller is designed to guarantee the stability. Ulti-mately, the effectiveness of the proposed method is corroborated through simulation experiments conducted on the quadcopter UAV.