Distributed Model-Free Adaptive Predictive Control for MIMO Multi-Agent Systems With Deception Attack

Abstract

This work explores the challenging problems of nonlinear dynamics, nonaffine structures, heterogeneous properties, and deception attack together and proposes a novel distributed model-free adaptive predictive control (DMFAPC) for multiple-input-multiple-output (MIMO) multi-agent systems (MASs). A dynamic linearization method is introduced to address the nonlinear heterogeneous dynamics which is transformed as the unknown parameters in the obtained linear data model. A radial basis function neural network is designed to detect the deception attack and to estimate the polluted output that is further used in the controller design to compensate for the effect. Then, the DMFAPC is designed by defining a new expanded distributed output with a stochastic factor introduced. The bounded convergence is proved by using the contraction mapping method and the effectiveness of the proposed DMFAPC is verified by simulation examples.