Latest revision as of 10:46, 4 May 2026

Abstract

For nonlinear multi-agent systems (MASs) vulnerable to stochastic disturbances and cyber-physical attacks on both sensors and actuators, this paper proposes an adaptive self-triggered consensus control framework based on neural networks. By using a decentralized leader-follower eventtriggered strategy, the method avoids Zeno behavior and drastically reduces communication overhead by updating local state estimates only at designated triggering instants. An adaptive mechanism compensates for actuator attacks, and a neural network is integrated to approximate unknown nonlinear dynamics, thereby improving robustness against malicious attacks and uncertainties. To ensure stability, a lower bound on inter-event times is derived, and practical consensus is demonstrated using Lyapunov-Krasovskii analysis. Both homogeneous and heterogeneous MASs’ numerical simulations confirm that the technique guarantees bounded state convergence and reduces the impact of attacks.OPEN ACCESS Received: 22/11/2025 Accepted: 24/12/2025

Document