Improved dynamic output feedback control for T-S fuzzy systems against hybrid cyber-attacks via neural network method

Abstract

This paper proposes a novel neural network-based dynamic output feedback controller (NN-DOFC) for nonlinear systems subject to hybrid cyber-attacks. The nonlinear terms are modeled using Takagi–Sugeno fuzzy inference rules, and the NN-DOFC is introduced to ensure that the closed-loop system achieves asymptotic stability while satisfying the $(X,Y,Z)$-dissipative property. The conventional DOFC’s gains are integrated as partial weights within the NN-DOFC framework, indicating that the traditional approach can be considered a specific instance of the proposed NN-DOFC. The three-layer fully connected feedforward neural network architecture extends the capabilities of the traditional DOFC. Additionally, a Bernoulli process is employed to model the hybrid cyber-attacks, including denial-of-service (DoS) and deception attacks. Then, the values of the controller are partially obtained by solving linear matrix inequalities and partially optimized using a genetic algorithm. Finally, by comparing the stabilization effects of the traditional DOFC with the proposed NN-DOFC, the effectiveness of the latter is demonstrated.