Abstract
This paper discusses an approximate realization of nonlinear H∞ state feedback controller based on neural networks. A three-layer neural network is constructed and is trained to satisfy Hamilton-Jacobi inequality, then the state feedback H∞ controller is designed using the network solution. It will be shown that the network learning problem for the solution of Hamilton-Jacobi inequality can be formulated as a maximum value function optimization problem. A nondifferentiable optimization technique is used to develop a learning algorithm. The efficiency of the proposed method is demonstrated by a numerical example.
