Decentralized monitoring of leader-follower networks of uncertain nonlinear systems

Efforts in this paper seek to develop a new method to monitor for undesirable performance in the general leaderfollower network structure of autonomous agents. Concepts from optimal control and adaptive dynamic programming (ADP) are used to develop a novel metric which networked agents with uncertain nonlinear dynamics use to monitor each other with decentralized communication. The developed approach uses a data-driven concurrent learning-based policy to identify agent dynamics and functions used to characterize optimality conditions, which are then used to check for compliance with specified performance criteria.