Observer-Based Fault Tolerant Control for a Class of Nonlinear Systems via Filter and Neural Network

Abstract

A filter and neural network (NN) based fault tolerant control (FTC) strategy is developed for a family of nonlinear systems expressed in strict feedback form in the event of unknown system dynamics and actuator failures. Specifically, adaptive neural network (ANN) is first utilized to facilitate the state observer design such that unmeasurable system states can be obtained. Note that ANN is only used when designing state observer instead of being used when designing controller. In our method, filter technique is introduced to construct virtual control inputs, which can not only reduce the adverse effects caused by ANN approximation errors and state estimation errors, but also deal with the expansion problem of the differential terms. Moreover, the fault tolerant tracking controller is designed by combining backstepping technique with the proposed NN with a novel weight updating law that is different from the above ANN. Theoretical analysis and simulation results demonstrate that the proposed FTC strategy can ensure that the tracking error converges to a small region of zero when there exist actuator faults and unknown system dynamics.