Dual-mode resilient model predictive control for cyber-physical systems against DoS attacks

In this article, a resilient model predictive control algorithm incorporating a dual-mode control strategy is proposed for a class of cyber-physical systems subject to state and input constraints, additive disturbances and denial-of-service (DoS) attacks. An adversary aims to disrupt the controller-to-actuator channel by deploying adversarial jamming signals. By exploring the prediction uncertainty between a nominal model and an actual model and utilizing a constraint tightening technique, a novel set of constraints is developed, which guarantees robust constraint satisfaction in open-loop mode caused by DoS attacks. A robust positive invariant set is designed, which allows the control mode to switch from rolling optimization to a state-feedback control law when the state enters its interior, thus saving computational resources. We explicitly determine a critical duration for DoS attacks and, through theoretical analysis, establish that when the duration of DoS attacks falls below this critical value, the proposed algorithm ensures guaranteed recursive feasibility and robust asymptotic stability. Finally, some comparisons are provided to illustrate the effectiveness of the proposed approach.