Optimal Output Feedback Control of Nonlinear Partially-Unknown Constrained-Input Systems Using Integral Reinforcement Learning

Abstract

The problem of output feedback optimal control of the partially-unknown nonlinear systems with constrained-input is investigated in this paper. Firstly, a neural network observer is proposed to estimate the unmeasurable system states. Secondly, synchronous integral reinforcement learning (SIRL) algorithm is used to solve the Hamilton–Jacobi–Bellman (HJB) equation associated with non-quadratic cost function and the optimal controller is obtained without knowing the system drift dynamics. This algorithm is implemented by a critic-actor structure and the novel weight update laws of the neural networks are designed and tuned simultaneously. Moreover, the weight estimation errors of all neural networks are proven uniformly ultimately bounded (UUB), and the stability of the whole closed-loop system is also guaranteed. Finally, two numerical simulation examples support the effectiveness of the proposed methods.