Optimal Finite-Time Tracking Control for a Class of Unknown Nonlinear System Based on Input-Output Data

Abstract

This paper appears an optimal tracking control method for completely unknown discrete-time nonlinear affine system. We make iterative adaptive dynamic programming (ADP) to approximately solve the Hamilton-Jacobi-Bellman equation by minimizing the finite-time performance index function. Based on input to output data, using model neural network as an identifier to construct the system input to output mapping, which is utilized to build the augmentation system. Then the action and critic neural network are utilized to approximate the virtual control and the corresponding performance index function, respectively. It proves that the estimation errors of the neural network are uniformly ultimately bounded. At last, an example is used to demonstrate the theoretical results and the performance of the proposed approach.