Observer-Assisted Model-Free Adaptive Predictive Control for Distributed Heterogeneous Multiagent Systems Against DoS Attacks

In this article, an observer-assisted model-free adaptive predictive control algorithm is designed for nonlinear heterogeneous multiagent systems (MASs) under Denial-of-Service (DoS) attacks. To begin with, consensus error of the system is transformed into an equivalent linear data model at the operating point by means of a dynamic linearization technique. Thereafter, in order to conserve network communication resources, a dynamic event-triggering mechanism is devised for the consensus error and an observer is incorporated to facilitate the acquisition of a time-varying parameter called “pseudo partial derivative (PPD)” in the linear data model. Subsequently, to mitigate impact of random DoS attacks, we introduce an attack compensation strategy that allows each agent’s output to swiftly revert to the neighborhood of leaderless consensus objective in DoS-free intervals. Finally, the PPD parameter estimates, observation errors, and consensus errors are rigorously proved to be bounded in mean using the contraction mapping principle. Simulation results on heterogeneous nonlinear MASs demonstrate that, compared with conventional event-triggered model-free control, the proposed approach improves the convergence rate by 33.82%, reduces steady-state consensus error by 93.80%, and decreases preconvergence fluctuations by 21.91%. Furthermore, it achieves a 64.00% faster recovery speed and a 95.98% higher recovery accuracy after DoS attacks. Meanwhile, the average communication load is reduced by 59.79%.