Adaptive control of nonstrict-feedback cyber-physical systems with unmodeled dynamics and false data injection attacks

This paper addresses the problem of adaptive neural control for a high-order nonlinear cyber-physical system in a non-strict feedback form, subject to unmodeled dynamics and false data injection (FDI) attacks on actuators. A dynamic signal is introduced to suppress the effects of unmodeled dynamics. Radial basis function neural networks (RBFNNs) are employed to approximate the uncertain nonlinearities and actuator attacks, while the norm of the RBFNN weight vector is estimated to reduce computational complexity. To handle unknown time-varying input gains arising from FDI attacks, a Nussbaum-type function is utilised. An adaptive backstepping control strategy is then developed to ensure that all signals in the closed-loop system remain bounded and the tracking error converges to a small neighbourhood of zero. Finally, two simulation examples are presented to demonstrate the effectiveness and robustness of the proposed control scheme under actuator attacks and system uncertainties.