Date of Award

1993

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

S. Sitharama Iyengar

Abstract

This project uses N-Body computer generated robots to investigate the fundamental problem of synergistic control of robots. This facility is important in an automated manufacturing environment and relates directly to the expanding field of study involving nonlinear systems. Experiments are performed utilizing this force controlled model to investigate nonlinear control problems using one of the techniques advanced. The principal results of this work are summarized: (1) A new simulation package is developed which uses Lagrange equations with multipliers to model two dimensional user defined robots. Three dimensionalization and parallelization are discussed and a dynamic optimization algorithm is introduced for regulating the size of the time increment. (2) A control package shell is developed which allows the user to construct and test control methodologies against this robot model; it also allows response time delays to be incorporated into the simulation. (3) A set of programs are developed as tools for investigating the synergism problem: neural networks, genetic algorithms, and a controller based on Newtonian physics. (4) Modified versions of genetic algorithms are investigated and faster convergence rates are obtained. (5) A reinterpretation of gradient decent is proffered and methods are found which speed backpropogation convergence in a realtime learning environment. (6) A controller is developed, based on Newton's second law, which allows us to determine the "effective mass" and to quantity and compensate for external forces. (7) Experiments similar to the broomstick and inverted pendulum are performed using the Newtonian control process. (8) An experiment is performed on a human torso robot.

Pages

99

Share

COinS