Robust and Resilient Model Predictive Control for Cyber-Physical Systems Against DoS Attacks




Dai, Yufan

Journal Title

Journal ISSN

Volume Title



With the development of Industrial 4.0, cyber-physical systems (CPSs) have been widely investigated due to their broad applications in a variety of areas. In a CPS, the cyber layer is integrated seamlessly with the physical components through a network-based structure, which dramatically alleviates the physical limitations at a low cost. However, great efficiency also comes with potential threats: The network-based structure is normally fragile and vulnerable to cyber attacks. These attacks can sabotage the elements in the system and tamper with the data, causing severe security problems, especially in the control system since the control system is the core infrastructure in most facilities. In this regard, model predictive control (MPC) stands out to be a promising solution to tackle attacks and ensure performance. Motivated by this fact, in this thesis, we focus on the robust and resilient MPC framework design and application against cyber attacks in CPSs. In chapter 2, a robust and resilient MPC scheme is proposed and utilized to drive an autonomous underwater vehicle (AUV) to track a predesigned trajectory. Nevertheless, the remote controller-to-actuator channel suffers randomly existing DoS attacks. Thus, a compensation strategy must be developed to mitigate the risk of the AUV going out of control. Thus, the packet transmission strategy is utilized in this work to construct a candidate control sequence at each sampling instant. By updating the sequence in the buffer every time the channel is not suffering attacks, the AUV can at least receive the control input torque to achieve its original control objective. Furthermore, the robustness constraint approach is also introduced in this work to deal with external disturbances. The effectiveness of the proposed method is verified by simulation results and its advantages are reflected through comparison study with the standard MPC approach. In chapter 3, a novel robust and resilient distributed MPC framework is proposed for the multi-agent CPS, in which all the communication channels among agents suffer randomly existing DoS attacks. Existing work only focuses on designing control strategies for the channels inside each agent (controller-to-actuator channels and sensor-to-controller channels), but neglects the influence of neighbor agents. To address this issue, a lengthened packet transmission strategy is proposed. By lengthening the predicted state sequence at each sampling instant based on the state-feedback control law, each agent in the CPS is able to receive the necessary information to steer its state to the small region around the equilibrium no matter when the attacks occur. A new type of robustness constraint is also designed to enlarge the region of attraction of this problem. Numerical simulation results for a four-ground-vehicle system are presented to illustrate the advantages of the proposed framework. In chapter 4, conclusions and potential future work are summarized and presented.



Cyber-physical systems, Model predictive control, Multi-agent systems, Trajectory tracking control, Denial-of-service attacks, Resilient control