Data-Driven Iterative Learning Security Consensus for Nonlinear Multiagent Systems With Fading Channels and Deception Attacks

Abstract

This work investigates the secure data-driven iterative learning control (ILC) problem for a kind of nonlinear discrete-time nonaffine multiagent systems under channel fading (CF) phenomenon and deception attack (DA). The stochastic fading behavior in the output channel is established as an independent Gaussian distribution model, the DA initiated by malicious attackers in the network damages the security of original data of each agent by injecting false data information. Relying solely on the incomplete output/intput data of every agent, the system model could be transformed into an equivalent data-driven form with adjacent-agent dynamic linearization (ADL) technology. And then the data-driven ILC algorithm gained through optimizing the two performance index functions makes the tracking error converges to a small neighborhood of zero in the sense of mathematical expectation. Finally, after rigorous theoretical analysis, the experiment confirms the practicability of the proposed algorithm.