Date of Award

8-2007

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Electrical and Computer Engineering

Advisor

Dawson, Darren M

Committee Member

Walker , Ian D

Committee Member

Wagner , John R

Committee Member

Burg , Timothy C

Abstract

This dissertation describes the design and implementation of various nonlinear control strategies for robot manipulators whose dynamic or kinematic models are uncertain. Chapter 2 describes the development of an adaptive task-space tracking controller for robot manipulators with uncertainty in the kinematic and dynamic models. The controller is developed based on the unit quaternion representation so that singularities associated with the otherwise commonly used three parameter representations are avoided. Experimental results for a planar application of the Barrett whole arm manipulator (WAM) are provided to illustrate the performance of the developed adaptive controller.
The controller developed in Chapter 2 requires the assumption that the manipulator models are linearly parameterizable. However there might be scenarios where the structure of the manipulator dynamic model itself is unknown due to difficulty in modeling. One such example is the continuum or hyper-redundant robot manipulator. These manipulators do not have rigid joints, hence, they are difficult to model and this leads to significant challenges in developing high-performance control algorithms. In Chapter 3, a joint level controller for continuum robots is described which utilizes a neural network feedforward component to compensate for dynamic uncertainties. Experimental results are provided to illustrate that the addition of the neural network feedforward component to the controller provides improved tracking performance.
While Chapter's 2 and 3 described two different joint controllers for robot manipulators, in Chapter 4 a controller is developed for the specific task of whole arm grasping using a kinematically redundant robot manipulator. The whole arm grasping control problem is broken down into two steps; first, a kinematic level path planner is designed which facilitates the encoding of both the end-effector position as well as the manipulators self-motion positioning information as a desired trajectory for the manipulator joints. Then, the controller described in Chapter 3, which provides asymptotic tracking of the encoded desired joint trajectory in the presence of dynamic uncertainties is utilized. Experimental results using the Barrett Whole Arm Manipulator are presented to demonstrate the validity of the approach.

Share

COinS