Robust output feedback fault‐tolerant control of non‐linear multi‐agent systems based on wavelet neural networks

Abstract

A robust output feedback active fault-tolerant leader-following controller for a class of non-linear multi-agent systems is presented. It is assumed that the states of the followers are not available; therefore, a local observer is constructed to estimate the states of each agent. In addition, the non-linear dynamics of agents may include uncertainties and the control input of the leader dynamics is unknown to all followers. Moreover, taking advantage of wavelet neural networks (WNNs), an online fault estimation scheme is developed which can effectively approximate the unknown actuator faults. The proposed decentralised observer-based robust cooperative controller is capable of compensating for the effects of unknown time-varying additive actuator faults, the model uncertainties, and the unknown input of the leader simultaneously. The stability analysis and convergence results that guarantee boundedness of all closed-loop signals are investigated via Lyapunov's direct method. To demonstrate the effectiveness of the proposed approach, a network of single-link manipulators is studied. As the results verify, the proposed WNN-based fault estimation scheme can properly approximate the unknown actuator faults, which results in efficient compensation in the fault-tolerant control design to achieve cooperative tracking objectives.