Design of resilient supervisory control by labeled Petri nets under external attacks

Abstract

The networked feedback control systems are susceptible to cyberattacks. This research focuses on the synthesis problem of a resilient supervisor under external attacks in DESs. An attacker can modify sensor information or re-enable actuator events, causing the system to reach predefined unsafe states. Each unsafe state is associated with a safety level, which indicates the level of risk or potential hazard related to the unsafe state. This paper aims to build a resilient supervisor that prevents the system from reaching the unsafe state when the number of attacks is less than the safety level of the unsafe state. A labeled Petri net is first used to model a plant, and a BRG is used to hide the unobservable or uncontrollable events in the system. Given a maximally permissive supervisor called a sub-supervisor before each attack, these sub-supervisors are connected via attacked events to create an automaton representing the resilient supervisor. After each attack, the resilient supervisor activates the corresponding sub-supervisor to ensure normal system operation despite the attacks. To make the resilient supervisor observe the occurrence of sensor attacks, a new algorithm is proposed by adding an auxiliary observable and uncontrollable self-loop. The proposed method in this paper has low computational effort and improves the system’s robustness.