This
project investigates the benefits of force feedback for virtual
reality training. Three groups of subjects received different
levels of training before completing a manual task, the construction
of a LEGO(tm) biplane model. One group trained on a Virtual
Building Block (VBB) simulation which emulates the real task
in a virtual environment, including haptic feedback. A second
group also trained on the VBB system, but without the force
feeback. The last group received no virtual reality training.
Completion times for these different groups in building the
actual biplane model in the real world were compared.
Purpose
Evaluate
the impact of force feedback on the effectiveness of virtual
reality training for a manual task performed in the real world.
Apparatus
The
Virtual Building Block System
The
system allows a user to manipulate building blocks within a
3-dimensional virtual environment. The distinctive feature
of this virtual reality simulation is that the subject cannot
only see the objects on the screen, he or she can actually 'feel'
them as they bump into each other, slide across one another,
and then snap into place.
-Excalibur - the haptic device
The user grabs the handle of Excalibur to move about within
the virtual environment. The device provides the virtual
world with the user's position and, in-turn, provides the user
with force feedback cues from virtual block interactions.
Excalibur has an approximately one foot cubic workspace and
up to 45 lbs of peak force output.
The
Virtual Building Block Software
At
the core of the software is a virtual coupling network, designed
to guarantee interaction between the human operator and the
virtual environment remains both crisp and stable. The
critical functions of device I/O, control, collision detection,
and object dynamics are performed on a Pentium IITM
PC at a rate of 1000 Hz. A second PC, equipped with OpenGLTM
acceleration, is dedicated solely to graphics. The two
computers communicate through a serial connection.
The
virtual building block system currently supports 10 different
types of LEGOTM-like pieces. Up to 50 blocks
can be simulated at once. The operator selects a piece
by moving a 3-D cursor inside of it using Excalibur and pushing
the left button on a cordless mouse held in his or her other
hand. Once a block is selected, the user has control of
it and can move it around in 3-D space. When the selected
piece collides with another block, the operator feels the impact
forces rendered through Excalibur.
The
operator can use keyboard commands to adjust the working view
right, left, up, down, in, or out. The virtual model can
also be flipped upside-down to work on the underside.
Multiple pieces can be clustered together and moved around at
once.
Experimental Design
General
Description
A total of 15 subjects undergo different
levels of training before building a 37 piece biplane model
from real LEGOTM blocks.
Dependent
Variable
The dependent variable is time to complete the real biplane
model. Each subject builds the model a total of 5 times.
Factors
(or Treatments)
The treatment in this study is level of training received before
building the real biplane model. There are three levels:
- Treatment
1 - virtual training with haptics. After receiving
a system familiarization, the subject watches a 4 minute
video on biplane construction. The subject then practices
building the biplane for 30 minutes on the virtual building
block system with force feedback.
- Treatment
2 - virtual training without haptics. After receiving
a system familiarization, the subject watches a 4 minute
video on biplane construction. The subject then practices
building the biplane for 30 minutes on the virtual building
block system without force feedback.
- Treatment
3 - no virtual training. The subject watches a 4 minute
video on biplane construction and directly proceeds to the
real task.
The test is
a between-subjects design, each subject receives only one of
the treatments. The 15 participants are separated into
3 groups of 5. A within-subjects design is impractical
for a training study such as this one since once the subject
has had the chance to build the real model 5 times, any further
training effect is insignificant.
Matched
Groups
A pre-test is used in this study to create three groups of subjects
with roughly equivalent ability to perform the manual task (biplane
construction). Before participating in the actual biplane
experiment, each of the 15 test subjects were asked to build
a 27 piece LEGOTM model of a hydrofoil boat 3 times.
Using their average time, they were then sorted from fastest
to slowest. The first 3 on this list was assigned to the
first matched group, the second three to the second, and so
on to get 5 matched groups. Each matched group of 3 was
then divided randomly among the 3 treatments.
