The Quest for Real-Time Virtual Human Control

Norman I. Badler
University of Pennsylvania

For over 30 years we have investigated methodologies to control and animate
virtual humans in real-time.  From our early work in inverse kinematics and
locomotion, to finite-state-machines, to methods adapted from motion analysis
and psychology, our quest has been to improve controllability, task animations,
and behavioral realism.  Our Parameterized Action Representation (PAR) links
animation control to higher level requirements such as executing instructions
and understanding Natural Language.  Our EMOTE work sought to add motion
qualities to vary a given animation. Recent work on eye movement and attention
exposes some of the subtle details needed for human simulation.  The next
generation of problems in computer animation will require tools for animating
large heterogeneous groups of individuals “living” their daily lives in a
complex environment.