Optimized Backstepping Consensus Control Using Reinforcement Learning of Observer-Critic-Actor Architecture Based on Neural Network for a Class of Nonlinear Strict-Feedback Multi-agent Systems

In this paper, a neural network-based adaptive optimized backstepping control scheme is developed, utilizing a reinforcement learning strategy, for a class of nonlinear strict-feedback multi-agent systems with unmeasured states. The neural network approximation-based reinforcement learning (RL) is performed under observer-critic-actor architecture. In the proposed optimized control, on the one hand, the state observer method can avoid to require the design constants making its characteristic polynomial Hurwitz; on the other hand, the proposed control method derives its RL updating laws from the negative gradient of a simple positive function related to the HJB equation, resulting in a significantly simpler algorithm. Meanwhile, it can also release two general conditions, known dynamic and persistence excitation, which are required in most of the RL-based optimal control.