On the Detection of Deception Attacks and Resilience in Turn-Based Command Governors for Dynamically Decoupled Systems

The distributed Command Governor (CG) approach is an effective solution that can be employed to supervise and coordinate a set of dynamically decoupled systems, connected via data links, to ensure local and global constraints satisfaction. Here the problem of attack detection is investigated under the assumption that the agents are subject to deception attacks. In this type of cyber-attack, typically an adversary can inject data into the communication link between agents, replacing legitimate commands with altered versions. This can render coupling constraints with neighbors infeasible, leading to constraint violations. Under mild assumptions, sufficient conditions for identifying attack-free scenarios are introduced. Building on this analysis, a decentralized CG scheme that adapts coupling constraints to account for estimation uncertainties, is proposed. Eventually, the effectiveness of the proposed supervision architecture is assessed via a simulative example involving a set of autonomous vehicles under local and connectivity keeping constraints.