Model-Based Actuator Fault Diagnosis in Multirotor UAVs*

Abstract

This paper presents actuator fault diagnosis for multirotor UAVs based on a cascade of a nonlinear Thau observer and a linearized Kalman filter. An example of actuator fault is when one or more rotors fail to deliver the required thrust to stabilize the UAVs. To simplify the presentation, we present a standard model for a quadcopter derived from Newton-Euler equation. The nonlinear model is locally Lips-chitz due to the cross and dot products between the angular and linear velocity vectors. The adaptive observer is designed assuming the angular velocity measurement is available, such that the model becomes linear time-varying (LTV). We show if the fault is constant then the observer goes to the actual value asymptotically, while if the fault is time-varying then the difference between the observer and the actual value will only be uniformly ultimately bounded. The proposed algorithm enables the system to detect, isolate, and estimate the magnitude of the faults. Numerical simulations show the proposed method able to estimate the state and the fault accurately.