Resilient MPC Under Severe Attacks on Both Forward and Feedback Communication Channels

This paper proposes a resilient model predictive control (MPC) strategy for constrained cyber-physical systems (CPSs) subject to disturbances and cyber attacks. The feedback sensor-controller (S-C) channel suffers from replay attack and the forward controller-actuator (C-A) channel suffers from false data injection (FDI) attack, and the defender has no prior information about the intruder. Considering that the abnormal behavior of intruder cannot be predicted, an expected one-step controllable set, and a series of minimally conservative constraints are developed to build attack detector. Two controllers are designed jointly based on infinite-horizon MPC to mitigate the negative effects caused by attack. Compared with the existing resilient control strategies, the attack model considered in this paper is less conservative, the resilient control structure is simpler, and it can avoid continuous channel refreshing caused by close-range attack. Robust constraint satisfaction, recursive feasibility and uniformly ultimate boundedness (UUB) are ensured for any admissible attack scenario and disturbance realization. Finally, simulations on a supply chain model show the efficacy of the algorithm.
Note to Practitioners
—With the wide application of wireless networks, vulnerabilities in the communication process can be easily exploited by intruders to launch malicious attacks. Resilient control can provide acceptable robustness and improve system safety. Additionally, many practical systems are constrained and disturbed, and MPC is one of the most effective methods for dealing with control problems in such systems. Its rolling optimization characteristics make it robust to disturbances. On the one hand, both the C-A channel and the S-C channel of the networked control system are vulnerable to attack. On the other hand, in a complex environment, the defender may not have prior information about the attacker, such as attack probability. Therefore, it is not practical to assume that only a single channel is attacked or that the algorithm relies on the attacker’s prior information. In this paper, a resilient MPC control structure is proposed to counter two types of attacks: replay attack and false data injection attack. One of its characteristics is that the attack model’s conservativeness is relatively low, and the structure is relatively simple, making the algorithm more applicable to actual needs. Another feature of the proposed approach is that it can be adapted directly to different types of attacks without modifying the overall resilient MPC architecture.