Event-Triggered Data-Driven Iterative Learning Control for Multiagent Systems With FDI Attacks

This work investigates the consensus learning control for heterogeneous nonlinear multiagent systems (MASs) under false data injection (FDI) attacks on the communication channels. An enhanced iterative dynamic linearization (EiDL) method is introduced to transform the nonlinear MAS into an equivalent linearization data model, where additional parameters are used to reflect the uncertainties of the MAS. Assume that the communication among agents is subject to a stochastic FDI attack which is modeled by a weighted sum of attacks for adjacent communication channels. Then, combining the event-triggering condition along the iterative direction, an event-triggered data-driven iterative learning control (ET-DDILC) is proposed where the attacked information is used in control law and parameter estimation law to counteract the impact of FDI attacks. The convergence is proven by introducing additional tools of mathematical expectations and matrix theory. Moreover, the proposed ET-DDILC is further extended to the MASs under iteration-switching topologies. Extensive simulation results verify that the proposed ET-DDILC can achieve a good control performance against injection attacks without using any model information while simultaneously saving system resources through the event-triggering mechanism.