MPC-based defense strategy for distributed networked control systems under DoS attacks.

In this paper, a distributed stochastic model predictive controller is designed for a networked control system subject to stochastic disturbances and denial-of-service (DoS) attacks. Each subsystem is constrained by actuator saturation and all of them are coupled via probabilistic constraints. By using the information on stochastic disturbances and considering the loss of data packet resulting from DoS attacks explicitly, an observer is designed to reconstruct the states and coupling probabilistic constraints are transformed into coupled deterministic constraints. By adopting the update strategy that only one subsystem is optimized at each time step, while others are setting feasible solutions as their control inputs, the constraints are further transformed into local deterministic constraints. To mitigate performance degradation due to DoS attacks, a performance constraint is introduced such that the optimal performance is not much worse than when the networks work perfectly. Constraints satisfaction, recursive feasibility and quadratic stability can be ensured despite the presence of DoS attacks. A numerical simulation is given to demonstrate the efficiency of the proposed algorithm.