Hamiltonian Critic Based Controllers for Stochastic Systems

点击翻译按钮获取中文摘要

9313946 Balakrishnan Artificial neural network based controllers for nonlinear systems will be studied int he project. Two main objectives of this project are to formulate neural network controllers based on adaptive critic designs and stochastic representations of uncertain nonlinear systems. Hamiltonian and its derivatives which occur in the formulation of optimal control problems are proposed in the adaptive critic design. A constant network configuration controller and an online-modified network configuration controllers are proposed based on neighboring optimal control theory. The stochastic representations of a plant are proposed to be based on filtering theory. Furthermore, classes of problems are described on which these network controllers are proposed to be investigated. It is hoped that the results of this study will substantially improve understanding of the adaptive critic based neuro controllers and the stochastic neural network representations of uncertain nonlinear systems. ***