Unknown-input-proportional-differential Observer-based Event-triggered Intrusion-tolerant Control for Human-in-the-loop Multi-agent Systems against Unconstrained Actuator and Sensor FDIAs

This paper investigates an intrusion-tolerant control problem for human-in-the-loop multi-agent systems (HILMASs) subjected to external disturbances and unconstrained actuator and sensor false data injection attacks (FDIAs) under directed graph. It is critical to emphasize that once a hacker gets to the control loop of the HILMAS, he can do whatever damage he wants, which means that the attack signal should be free of any constraints. Two main unconstrained attacks, i.e., unbounded FDIAs and variable-frequency FDIAs, are hard to accurately estimate and defend since the widely adopted constraints on the existing FDIAs such as the bounded or/and the bounded first-order derivative have been removed. To tackle this challenging obstacle, a novel unknown-input-proportional-differential observer (UIPDO) is developed to not only reconstruct the follower agents’ states as well as unconstrained actuator and sensor FDIAs simultaneously, but also avoid the decrease of estimation accuracy caused by measurement deviation. It should be noted that this measurement deviation may be extremely large as it is caused by unconstrained sensor FDIAs, which renders the traditional observer ineffective in providing reliable and accurate estimates of the system states and unconstrained actuator and sensor FDIAs. Then, a novel UIPDO-based intrusion-tolerant control strategy without requiring boundedness of the first-order derivatives of the FDIAs as in existing literature is proposed. Furthermore, an adaptive Zeno-free event-triggered mechanism (ETM) solely relying local state information is developed to reduce the communication burden. Finally, the numerical simulation is provided to verify the merits and effectiveness of the developed methodology.