Reinforcement learning and game-based optimal output consensus control for higher-order multi-agent systems with unknown dead-zone inputs

This article focuses on a type of higher-order nonlinear multi-agent systems (MASs) subject to unknown dead-zone inputs and investigates the leader-follower practical output consensus control problem for such system. A reinforcement learning (RL) and backstepping methods-based controller is developed to ensure that the differential graphical game-based performance function achieves the minimum value, where the actor and the critic neural networks (NNs) in the RL framework are respectively employed to execute control policy and assess control performance. By utilizing the devised optimal controller, a practical output consensus criterion is obtained for the higher-order nonlinear MAS and the resulting distributed control policies constitute a Nash equilibrium. A numerical simulation is finally presented to verify the effectiveness of the devised optimal control scheme.