Multivariable Adaptive Super-Twisting Sliding Mode Resilient Control for Uncertain Nonlinear CPSs Against Actuator and Sensor Attacks

This paper proposes an adaptive super-twisting sliding mode resilient control method for uncertain nonlinear cyber-physical systems (CPSs) subject to actuator and sensor attacks. Two super-twisting algorithms with adaptive gains are designed based on the compromised system states to counteract the effect of cyber-attacks to ensure the system states converge to a small region near zero. Furthermore, the gain overestimation problem of the sliding mode resilient controller is also discussed. Finally, a formal analysis of the closed-loop system is derived using Lyapunov function techniques. Simulation results on a two-link manipulator system validate the effectiveness of the proposed resilient control strategy Note to Practitioners-The motivation of this paper is to address the secure issues of the nonlinear cyber-physical systems, including applications in robotic manipulators, vehicle systems, and power systems. Since some devices of CPSs, such as sensor and controller components, communicate via networked channels in practice, the original system states and control signals of CPSs may be tampered by cyber-attacks, leading to unavailable system states and severe uncertainties with unknown bounds. Although the super-twisting sliding mode control (ST-SMC) method is an effective tool for dealing with uncertainties, the control gains may be overestimated due to the intermittent nature of cyber attacks, resulting in severe chattering caused by the discontinuous term of the ST-SMC method. Furthermore, it is extremely challenging when only compromised system states are available for the controller, potentially generating false commands and destroying the entire system. Therefore, based on Lyapunov theory, a novel multivariable adaptive sliding mode resilient control method with two adaptive super-twisting algorithms is proposed to mitigate the effects caused by cyber-attacks. The effectiveness of the proposed method is validated using a two-link manipulator. Note that the proposed method has the potential for broader engineering applications in real-world scenarios.