Document Type

Dissertation

Date of Award

1995

Degree Name

Doctor of Engineering (DEng)

Department

Electrical and Computer Engineering

First Advisor

Nikolaos G. Bourbakis

Second Advisor

George J. Klir

Third Advisor

N. Eva Wu

Subject Heading(s)

Robots -- Control systems; Robotics; Academic theses

Abstract

Neural networks and fuzzy logic are combined into a hierarchical structure capable of planning, diagnosis, and control for a redundant, nonlinear robotic system in a real world scenario. Throughout this work levels of this overall approach are demonstrated for a redundant robot and hand combination as it is commanded to approach, grasp, and successfully manipulate objects for a wheelchair-bound user in a crowded, unpredictable environment. Four levels of hierarchy are developed and demonstrated, from the lowest level upward: diagnostic individual motor control, optimal redundant joint allocation for trajectory planning, grasp planning with tip and slip control, and high level task planning for multiple arms and manipulated objects. Given the expectations of the user and of the constantly changing nature of processes, the robot hierarchy learns from its experiences in order to more efficiently execute the next related task, and allocate this knowledge to the appropriate levels of planning and control. The above approaches are then extended to automotive and space applications.

Included in

Engineering Commons

Share

COinS