Date of Award

Spring 1996

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Mechanical Engineering - (Ph.D.)

Department

Mechanical Engineering

First Advisor

M. C. Leu

Second Advisor

Bernard Friedland

Third Advisor

Denis L. Blackmore

Fourth Advisor

Nouri Levy

Fifth Advisor

Zhiming Ji

Abstract

Cell mapping methods can generate global optimal controls for nonlinear dynamic systems. However, the implementation of this approach for on-line control is difficult due to its use of a table-based controller which requires a large amount of memory.

In this dissertation, a cell mapping based systematic method is developed to construct a fuzzy controller to replace the table-based controller for global optimal control of dynamic systems. The method consists of four steps: 1) An optimal control table is generated using a cell mapping algorithm. 2) The cells are linked into trajectories to characterize the optimal dynamic evolution of various cell. 3) The trajectories are classified into groups of similar trajectories so as to represent global optimal controls for the entire cell space with a number of representative trajectory groups. 4) Based on the input-output relations of these trajectories groups, which represent the states and controls of the corresponding cells, a set of fuzzy rules are generated for a fuzzy controller.

With the method of grouping trajectories developed in this dissertation, controls for the entire space can be expressed corresponding to trajectory groups instead of cells with a reduced number of trajectory groups. This reduces the complexity of constructing the global fuzzy controller compared with developing rules based on the control data of all the cells. The developed method is applied to the ship navigation and car parking problems to show the applicability of the method to two and three dimensional problems.

Share

COinS