The JEDI - Training

1
The JEDI - Training

• A Playful Evaluation of
• Interactive Head-Mounted
• Augmented Reality Display Systems

Christopher S. Oezbek, Björn Giesler and Rüdiger
Dillmann Institute for Computer Design and Fault
Tolerance (IRF) Chair for Industrial Applications
of Informatics and Microsystems
(IAIM) Universität Karlsruhe (TH), Germany
Electronic Imaging Conference Thursday, 22nd
January 2004 San Jose, CA
2
Background (1/2)

• Augmented Reality (AR) is the concept of adding
  3D virtual elements to the real world.
• This is in contrast to simulation or virtual
  environments where a real world does not exist.
  AR therefore requires
• Tracking of the position of real-world elements
  of interest
• Registration of the user in relation to the
  virtual world
• Furthermore, interactive AR does require easy and
  precise control of a set of virtual elements by
  the potential user. Some definitions of AR
  require interactivity as a mandatory
  characteristic.

3
Background (2/2)

• Although other senses like hearing, touching, and
  even smelling are included in this definition,
  the main focus of AR research is on the visual
  sense so far.
• Besides the limited exception of projection
  display systems, visual augmentation is most
  commonly achieved by using head-mounted displays.
  
• Two types of head-mounted displays have been
  realized and are available
• Optical See Through (OST)
• Video See Through (VST)

4
Optical See Through (OST)

• Using a partially transmissive mirror, the real
  world image is combined with virtual pictures
  generated in the HMD.
• Since no input is generated by the head-mounted
  display, head tracking needs to be done using an
  extra device.
• Advantages
• Real-world directly visible (Resolution, delay,
  eye offset)
• Safety
• System used in Jedi-Training
• Sony Glasstron
• Stationary NDI Polaris infrared tracker

5
Video See Through (VST)

• Video cameras capture the real world. Those
  pictures are combined with virtual elements and
  displayed to the user.
• Since input is gathered from the surrounding
  world, no special tracking device needs to be
  attached.
• Advantages
• Better composition strategies (opacity, delay)
• No registration required
• System used in Jedi-Training
• Trivisio ARVision 3D-HMD
• ARtoolkit image based marker tracking

6
Goals of the Experiment

• Compare both systems with respect to
• Performance and Effectiveness
• 3D feeling
• Tracking
• Comfort
• Acceptance
• Effect of preexisting knowledge

7
Methods (1/2)

• To achieve the named goals, a game was introduced
  resembling the laser light saber training from
  the film Star Wars.
• A small spherical remote robot was displayed in
  the users field of view. Moving randomly inside
  a predefined area, the drone fired laser rays
  against the player who had to fend them off with
  his light saber.
• For each succesful defend the player scored one
  point and the session ended after five
  unsuccessful defends.

8
A little Mock-up
9
Methods (2/2)

• Players were divided into two groups
• Group A started with OST and after three
  sessions
• crossed over to the VST system for another
  three sessions.
• Group B started with VST and moved on to OST
• After each run each player was asked to enter his
  name using a conventional keyboard and scores
  were recorded.
• A questionnaire on a scale of 0 - 10 was deployed
  afterwards to gather subjective data.

10
Study Population

• 31 CS students from University Karlsruhe
• 27 male, 4 female
• 13 participants wore spectacles
• 15 participants started with OST
• 16 participants started with VST

11
Study Population
N 31
n.s.
n.s.
n.s.
n.s.
12
Results
N 31
n.s.
plt0,05
plt0,05
n.s.
Initial 3 games The learning periode
13
Results
N 31
Group B
plt0.05
plt0.01
Score
n.s.
plt0.05
plt0.01
n.s.
Initial 3 runs Learning effect
Final 3 games Experienced level
14
Results
N 31
plt0,05
n.s.
plt0,05
plt0,05
plt0,05
15
Results
N 31
plt0,05
plt0,05
plt0,05
plt0,05
plt0,05
16
Conclusions

• The optical see through system achieves higher
  scores, is more comfortable to use, and more
  accepted than the VST solution.
• No measure of preexisting knowledge/skill
  correlated with score results.
• Image based tracking via the ARtoolkit should not
  be used for object tracking.
• High framerates and low delay are the most
  pressing issues for VST solutions.
• With forthcoming improvements comparative testing
  will become more neccesary.
• From the perspective of today both systems have
  the potential to become the gold standard.

17
Thank you