Leader-Follower Formations Subject to False Data Injections: A Resilient Distributed Model Predictive Approach

In this paper, resilience issues for platoons of autonomous agents are addressed when false data injections affect the information exchanged among the neighbors via a communication medium. A distributed model predictive control scheme is used for dealing with the overall regulation task. Conversely, the core of this study relies on the design of an efficient anomaly detector and viable attack countermeasures. In particular, it is formally proven that the proposed device is capable to uncover in finite time malicious actions by simple set-containment set-membership conditions arising from the concept of
$k$
- step ahead state predictions convex sets. Moreover, the attack countermeasures have a twofold nature: the first one is conceived by exploiting feasibility arguments of the model predictive philosophy; while the second resilient operation takes inspiration from rejuvenation ideas by leading to safe splitting and/or queuing the initial multi-agent formation.