Optimal Control of a Class of Nonlinear Systems Using Radial Basis Function Neural Networks

This paper presents an online optimal control technique for a class of nonlinear systems. The technique is based on approximating the solution to the corresponding generalized Hamilton-Jacobi-Bellman (GHJB) equation for optimal control using radial basis function neural networks (RBFNN). The GHJB equation is solved by adjusting the parameters (weights and centers) of RBFNN online. The proposed optimal control algorithm provides good accuracy and numerical examples illustrate the merits of the proposed approach.