Pattern Recognition for Imitating Human Behavior

                       Christian Thurau

                   University of Bielefeld, Germany

http://www.techfak.uni-bielefeld.de/ags/ai/members/homes/cthurau.html

Abstract:

Skill acquisition in robots and simulated agents has been a topic of
increasing popularity throughout the last years. Despite impressive
progress, autonomous behavior at a level of animals and humans are not
yet replicated by machines. Especially when a complex environment
demands versatile, goal-oriented behavior, current artificial systems
show shortcomings. Consider for instance modern 3D computer games.
Despite their key role for more emersive game experience, surprisingly
little effort was made towards new techniques for life-like behavior
creation in artificial characters.

Modern interactive computer games provide the ability to record
complex data on human behavior, offering a variety of interesting
challenges to the pattern-recognition community. Such recordings often
represent a mixture of long-term strategy, mid-term tactics and
short-term reactions, in addition to the more low-level details of the
player's movements. In this talk we we approach the goal of behavior
acquisition in artificial agents by means of machine learning and
pattern recognition. We assume the behavior of game characters to be a
function that maps a world state on a reaction. Various methods are
elaborated to extract behaviors operating on different timescales
(long-term, mid-term, and short-term) from behavior recordings. For
instance, a Bayesian approach maps world-states on a discrete set of
movement prototypes to imitate context dependent action sequences. For
experimental validation we use a commercial 3D game environment. The
experimental results show that human behavior can indeed be
bootstrapped from basic behavioral observation data. Moreover, a
survey indicates that the synthesized behaviors are perceived as
humanlike.