Adaptive Critic Based Neurocontrol for Distributed Parameter Systems

点击翻译按钮获取中文摘要

9976588BalakrishnanThis project will extend the adaptive critic designs in neural networks to the control of distributed parameter systems. One of the strengths of the adaptive critic design is that it presents a unified learning based approach regardless of the problem structure-whether it is linear or nonlinear. This research has two major objectives: the first is to develop two adaptive critic designs for the control of distributed parameter systems. The first design based on approximate dynamic programming outputs optimal return function and the derivatives of the cost with respect to the system states. The second critic design outputs the Hamiltonian and its derivatives with respect to the states and control. The second major objective is to analyze performance of these neuro-controllers in controlling parabolic (and hyperbolic) and elliptic systems. These types of systems represent distributed parameter control problems in MEMS, active flow control, and active vibration control, ocean mining for minerals, power systems, etc. We will consider regulator problems AND finite horizon problems. The stability and convergence of the adaptive critic designs will form part of the study. He plans investigate the effects of imperfect modeling which is a crucial issue in any controller synthesis for distributed parameter systems. The significance of this research lies in the fact that it combines the human problem solving (brain like intelligence) philosophy embedded in the adaptive critic design with classical mathematics to solve some very important and current complex problems in Engineering.***