Reconfiguration Control Method for Faulty Actuator on UAV

Abstract

In this paper, a fault-tolerant control (FTC) scheme is improved to address the locking and loss of effectiveness of a fixed-wing UAV actuator. The FTC scheme proposes a novel way of combining continuous and discrete framework to address both continuous estimation and control performance considerations, together with discrete decision making aspects. An augmented Kalman filter estimates the state and the fault parameters, and the solving of Partially Observable Markov Decision Process (POMDP) decides the feedback controller to be used. This approach is improved to reconfigure the UAV during a landing approach when an actuator becomes faulty. The improved approach allows for taking into account faults on a trimmed aircraft, generalizes the method for defining fault modes and associated controllers and allows for taking into account different time steps for the continuous and discrete components of the FTC. The efficiency of this approach is demonstrated by obtaining a $25 \%$ gain on the sum of tracking errors with respect to the nominal controller on a longitudinal model of the Altimum 4.