Resilient Containment Control of Multi-Agent Networks in Adversarial Environment

We study the problem of containment control in an adversarial environment, where some of the agents may be adversarial. We identify the security issue of a containment control protocol in such adversarial environment. We propose a resilient containment control protocol that ensures the state of each non-adversarial (normal) follower agent converges to the convex hull spanned by the states of the normal leader agents. Specifically, our protocol is based on the method of resilient convex combination and works for agents with a vector state. For multi-agent networks consisting of one follower agent and multiple leaders, we provide a sufficient condition on the communication topology that guarantees the success of our proposed protocol. In addition, we provide a set of numerical simulations to demonstrate the success of our protocol and verify our theoretical results.