Resilient Neural-Variable-Structure Consensus Control for Nonlinear MASs with Singular Input Gain Under DoS Attacks

Abstract: This paper proposes a reliable learning-based adaptive control framework for nonlinear multi-agent systems (MASs) subject to Denial-of-Service (DoS) attacks and singular control gains, two critical challenges in cyber-physical systems. A neural-variable-structure adaptive controller is developed to achieve leader-follower consensus while ensuring robustness to external disturbances and adaptability to unknown nonlinear dynamics. A reliability-assessment rule is introduced to detect communication loss during DoS attacks, upon which a switched control mechanism is activated to preserve closed-loop stability and performance. Unlike existing resilient MAS control methods, the proposed strategy explicitly accommodates singular control gains and does not rely on restrictive assumptions such as Lipschitz continuity or prior bounds on nonlinearities. To the authors' knowledge, this is the first work to integrate neural learning, variable-structure robustness, and reliability-based switching into a unified consensus-tracking control architecture for heterogeneous nonlinear MASs with singular input gains under DoS attacks. Lyapunov-based analysis establishes uniform ultimate boundedness of all closed-loop signals, and Matlab/Simulink simulations on a connected automated vehicle platoon demonstrate the method's effectiveness and resilience.