Data-Driven Control for Nonlinear Multi-Agent Systems Using Reinforcement Learning

In this paper, a new reinforcement learning-based model-free adaptive control algorithm is introduced for discrete-time nonlinear multi-agent systems with unknown dynamics, while the equivalent dynamic linearization algorithm is applied to design the optimal controller. The strategy for Q-Learning and the actor-critic neural network are specifically redesigned to achieve consensus control in multi-agent systems. The proposed reinforcement learning algorithm can adjust the dynamic linearization parameters in real-time only based on input and output data. The stability of the closed-loop system is proven by Lyapunov theorem. Furthermore, the method’s effectiveness is verified by a numerical simulation.