NN-based compensation control for uncertain delayed singular piecewise homogeneous jump systems with deception attacks and time-varying transition probabilities

Abstract

This work investigates neural network (NN)-based compensation control for uncertain delayed discrete singular piecewise homogeneous Markov jump systems (PHMJSs) under deception attacks. Considering the unmeasurable system states and limited network bandwidth, the states of the discrete singular PHMJS are estimated via utilizing a state observer and communication resources are saved through an improved event-triggered mechanism (ETM). For Markov chains, time-varying transition probabilities (TPs) resulting from environmental changes and external disturbances are considered to be piecewise homogeneous, whose stochastic variations are regulated by a higher-level transition probability (HTP) matrix. Moreover, in order to alleviate the adverse impact arising from deception attacks over the controller-actuator transmission channel, the NN technique is employed to approximate malicious attacks. Then, an event-triggered compensation feedback controller based on the reconstructed deception attacks is devised to compensate for the negative effect of deception attacks on systems. By taking advantage of singular value decomposition technique and double-mode-dependent Lyapunov–Krasovskii (L-K) functional, fresh conditions of the regularity, causality and boundedness in probability for discrete singular PHMJSs are obtained under the framework of linear matrix inequalities (LMIs). Finally, DC motor is provided to verify the feasibility of the proposed NN and ETM-based compensation control approach.