Classical and neural net control and identification of non-linear systems with application to the two-joint inverted pendulum control problem
Metadata[+] Show full item record
[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] A two joint inverted pendulum control system is an example of a highly nonlinear dynamic system which is investigated in this thesis in terms of control and system identification. These dual terms are very important in the area of adaptive automatic control in which parameter changes have a crucial effect on the performance of the controlled plant. The plant to be controlled consists of two pendulums that are to be maintained close to their vertical unstable equilibrium positions by applying a force on the cart. This thesis investigates classical optimal control techniques such as linear quadratic control applied for controlling the linearized and the nonlinear models of this two joint pendulum control system. It also investigates parameter estimation techniques using optimal Kalman filters. System identification, is often identified as an important step in the design and analysis of controllers for both linear and nonlinear plants. Therefore, linear model prediction techniques such as Box Jenkins and nonlinear approximation using neural networks are also investigated. The application of neural networks for system identification has provided interesting and useful results, which demonstrates the benefits of neural networks in application to control, and based on the results obtained in this investigation, the limitations in the use of neural networks to adaptive control have been observed.
Access is limited to the campuses of the University of Missouri.