UCPP: http://wiki.beyondunreal.com/wiki/UCPP
UsUnit: http://wiki.beyondunreal.com/wiki/UsUnit
Posted March 11, 2003, updated April 29, 2003
(New! On June 20
Neil Heffernan was notified
that the Army will be funding this project. I will be looking for MQP,
Independent Study Courses, and Graduate Student’s Theses/Disserations. There are other funding opportunities for
this sort of work like this one.)
I am offering an MQP related to adding an intelligent tutoring system to a
"first person shooter" sort of game. This project
could even start in D-term or maybe as a PreQ-P. We could run the first
term as an ISP. For grad students, this would make an excellent masters
thesis topic. As part of my research, I
just wrote a grant proposal to the Army in response to their request for
proposals.
Students will need to take some public domain first person shooter engine,
modify it to fit the Army's needs, build a expert system in JESS that knows how
to "play the game", turn that expert system into a tutoring systems,
and finally hook the game and the expert system together. This last step
will require us to put the right hooks into the simulation system.
Students that have done game development might be particularly interested.
If you are interested, please contact with me a resume, transcript, a letter indicating
why you would do a good job at this MQP. Send these materials, all
together, either electronically or snail mail. Students should probably
have taken, or are taking, Artificial Intelligence. This project will
involve a great deal of programming and should not be undertaken lightly.
Please tell me about the other large scale programming tasks you have
undertaken. The few (2-4) students that seem most likely to succeed will
be selected.
Try this link to where I found this
ARMY03-T01
TITLE: The Virtual Observer/Controller (O/C) --- Intelligent Coaching in
Dismounted Warrior Simulations
TECHNOLOGY AREAS: Human Systems
ACQUISITION PROGRAM: USAIS, Combined Arms Tactics
Directorate
Objective:
Develop intelligent, automated coaching and feedback for training
dismounted small unit leaders and teams within a collective virtual
simulation/computer gaming environment.
The intent is to merge two training technologies – intelligent tutoring
engines for individual skill training and virtual/gaming simulations for
small-unit, dismounted operations. A
synthetic, intelligent observer/controller (O/C) shall be created within
simulations to perform the real-time coaching and feedback functions similar to
those functions executed by actual O/Cs or unit leaders during field exercises
within a unit or at the Army’s Combat Training Centers.
Description: The
Army has invested considerable resources in developing simulations for
small-unit dismounted warrior operations.
Recent advances in technology have provided the necessary resolution and
detail in the terrain needed for dismounted operations. In addition, considerable progress has been
made in realistically displaying human behavior. Nonetheless, the training effectiveness and
value of such simulations continue to reside in the expertise of the individual
trainer. This trainer establishes the
rules for the exercises, watches how the unit executes a scenario, and then
provides feedback. The trainer typically
decides which exercise segments to replay and what results to display. In live, field exercises, the trainer or
observer/controller (O/C) not only provides the after-exercise feedback, but
also coaches during the exercise, to include stopping the exercise if
necessary. The purpose of the proposed
effort is to integrate intelligent tutor computer engines used for individual
skills with collective simulations of small-unit dismounted operations in order
to embed the warfighting expertise of “live expert coaches” in the
simulation. The merger of these
technologies would mean that the training value of the simulation would not be
as highly dependent on the training and military expertise of a single
trainer. Also by incorporating such
expertise there is some guarantee that the simulation exercise will result in
the desired training objective, not simply be an event or a “game.” Dismounted warrior operations also require
that intelligent tutor engines be modified to accommodate the “unbounded”
nature of such operations, where there can be multiple means of accomplishing
the mission, rather than just one solution.
Phase I:
Phase I shall consist of a front-end analysis to determine the missions
or collective tasks to be simulated, and identification of suitable measures of
performance and effectiveness. The tasks
shall be basic, dismounted Infantry unit collective tasks, that represent
differing degrees of uncertainty regarding how the tasks are best planned and
executed. The task domain shall be at
the small-unit level. An approach shall
be developed for modifying current intelligent tutoring engines to incorporate
feedback at the collective level as well as feedback to individuals. In addition, the engine shall be modified to
provide feedback that accommodates multiple solutions to a problem. Requirements for providing coaching during an
exercise and feedback following an exercise shall be determined. The performance data to be collected and
stored during the simulation shall be determined and a means of transferring it
to the intelligent engine identified.
Virtual scenarios that accommodate the tasks to be trained shall be
identified/modified. Subject matter
experts shall be interviewed and training and doctrine manuals reviewed to
determine the knowledge base required by O/Cs for at least one of the selected
tasks. The proposed solutions for
creating a synthetic intelligent O/C within a small-unit dismounted warrior
virtual/gaming simulation environment shall be documented in a Phase I
report. The report shall also include
findings from the front-end analysis of the collective tasks to train,
documentation of simulations available for these tasks, measures of
performance, simulation database requirements, the expert knowledge database
required for the selected task(s), and modifications required to intelligent
tutoring technology.
Phase II: In Phase II, the synthetic intelligent O/C
virtual training/coaching simulations agreed upon in Phase I shall be
developed. A milestone schedule for
generating the “intelligent O/C engine,” determining the individual and
collective behaviors to assess, and merging this capability with virtual
dismounted simulations and scenarios shall be established. Three dismounted warrior mission/tasks shall
be used. Assessments of the
effectiveness of the training products as well as reactions by soldiers and
leaders shall be conducted simultaneously with product development. An assessment plan shall be written for
review and approval. A report describing
these findings as well as the algorithms for the intelligent virtual O/C shall
be produced.
Phase III: The Department of Defense and many
corporations now use virtual simulations for leader and team training. The training capabilities described in this
proposal apply to these virtual multi-person simulation capabilities. Incorporating a synthetic trainer that
provides real-time coaching and performance feedback will result in enhanced
team training, and a capability that is less reliant on quality
instructors. In addition, a synthetic
trainer that adapts to multiple, yet suitable solutions for accomplishing
mission objectives developed by teams greatly expands the settings in which
intelligent agent technologies can be applied.
REFERENCES:
Brown, B. Wilkinson, S., Nordyke, J., Riede, D.,
Huyssoon, S., Aguilar, D., Wonsewitz, R., & Meliza, L. (1997).
Developing an automated training analysis and feedback system for tank
platoons (Research Report 1708).
Ericsson, K. A., & Smith, J. (Eds.). (1991).
Toward a general theory of expertise:
Prospects and limits.
Gillis, P. D., Hursh, S., Guest, M. A., Sweetman,
B., & Ehrlich, J. A. (2000). Cognitive behaviors for computer generated
forces (ARI Technical Report 1113).
Pleban, R. J., Eakin, D. E., Salter, M. S., &
Matthews, M. D. (2001). Training and assessment of decision-making
skills in virtual environments (ARI Report 1767).
Polson, M. C., & Richardson, J. J. (Eds.).
(1988). Foundations of intelligent
tutoring systems.
Shute, V.J., Lajoie, S. P. & Gluck, K. A. (2000).
Individualized and group approaches to training. In S. Tobias & J. D. Fletcher (Eds.),
Training and retrainng: A handbook for business, industry, government and the
military (pp. 171-207).
Singer, M. J., Grant, S., Commarford, P. M., Kring,
J. P., & Zavod, M. (2001). Team performance in distributed virtual
environments (ARI Technical Report 1118).
Wisher, R. A., MacPherson, D. H., Abramson, L. J.,
Thorndon, D. M., & Dees, J. J.
(2001). The virtual sand
table: Intelligent tutoring for field
artillery training (ARI Research Report 1768).
KEYWORDS: Intelligent tutors, virtual simulations,
computer gaming, dismounted operations, team training, intelligent feedback,
after action reviews
TPOC:
Dr. Scott E. Graham
Fax:
(706) 545-4618