Predefined-Time Distributed Optimal Control for Nonlinear Multiagent Systems via Reinforcement Learning

This paper investigates the predefined-time adaptive distributed optimal control problem for nonlinear multiagent systems with unknown dynamics and actuator saturation. A novel predefined-time distributed optimal controller is formulated by using neural network and differential graphical game. To solve the predefined-time optimal control design problem, a neural network reinforcement learning algorithm is developed to learn the solutions of the optimal controller and Hamilton-Jacobi-Bellman equation, where the updating laws of the critic network weights are designed via current and historical data. It is proved the controlled nonlinear multiagent systems are predefined-time stable and achieve the Nash equilibrium within a finite settling time. Finally the computer simulation results illustrate the effectiveness of the proposed distributed optimal control scheme.