|
|
An approach to human performance modeling for semi-autonomous agents in virtual environments
Dr. Yiannis Papelis
Monday, February 12, 2007
2:00PM ~ 3:00PM, Harris Center-125
Abstract
Modeling realistic human-like autonomous agents in immersive virtual
environments requires capturing the stochastic nature of human behavior.
Deterministic behavior that shows no variation or unpredictability is
unnatural and typically attributed to artificial entities, hence the term
robot-like. At the same time, it is often the case that the motives for
using immersive virtual environments, such as in research and training,
require and often depend on repeatability and determinism of interactions
between the human participant and the autonomous agents. The talk will
overview work in semi-autonomous agent modeling that attempts to address the
contradictory aspects of randomized yet deterministic behavior in immersive
virtual environments. The talk will also discuss the use of hidden markov
models (HMMs) for two purposes. First, as modeling tools HMMs can be
trained by real data and very effectively capture the stochastic nature of
human behavior. Second, as performance evaluators, HMMs can recognize
patterns of actions which in turn can be used to assess performance against
exemplars of behavior. The talk will probe the relation between the
stochastic nature of autonomous behavior and performance assessment of human
behavior in virtual and live environments.
Short Bio
Yiannis Papelis has a B.S. degree from Southern Illinois University, a M.S
from Purdue University and a Ph.D. from the University of Iowa, all in
Electrical & Computer engineering. Before joining UCF as a visiting
Assistant Professor, he held positions as R&D project leader and research
scientist at the University of Iowa. His prior work focused on immersive
virtual environments and driving simulation in particular. He has been the
principal or co-principal investigator for over 35 applied research projects
spanning work in immersive driving simulation, vehicle safety system
evaluation and driver performance. Highlights of this activity include
training systems that are currently used to train thousands of Swiss
soldiers every year, simulators used to research effects of drugs and
in-vehicle device distraction on drivers, research simulators used to design
advanced agricultural equipment, and findings that have been used by the
department of transportation to support a new law that if instituted, is
anticipated to reduce traffic accidents and save up to 10,000 lives each
year.
|
|