Virtual Humans:
Behaviors and Physics, Acting and Reacting

July 20, 1998

Norman I. Badler (Chair) -- Director, HMS Center, University of Pennsylvania
Armin Bruderlin -- Sony Pictures Imageworks
Athomas Goldberg -- Media Research Laboratory, New York University
Nadia Magnenat-Thalmann -- Miralab, University of Geneva
Dimitris Metaxas -- HMS Center, University of Pennsylvania
Ken Perlin -- Director, Media Research Laboratory, New York University

Last update, March 5, 1998

This course is intended to demonstrate the state-of-the-art in motion synthesis for virtual humans. It covers the essential elements for interactive and real-time animation via behavior, physics, signal processing, scripting, live motion capture, and reactive techniques.

This course is intended to demonstrate the state-of-the-art in interactive, networked, and real-time motion and behavior synthesis for Virtual Humans. Applications of such human animation include simulation and training, intelligent agents, avatars, and realistic motion synthesis for entertainment characters. The course covers the essential elements for synthetic animation via kinematics, inverse kinematics, dynamics, optimal and adaptive control theory, motion signal processing, motion texture synthesis, and video-based motion capture. Current progress in the MPEG-4 Synthetic/Natural Hybrid Coding group will be demonstrated, integrating real-time sound, face, and body animation. In addition, the course covers the higher level behaviors which impart a sense of decision-making and reactivity to make Virtual Humans appear to act more human. These techniques include collision avoidance, postural transitions, parallel finite-state machine controls, intelligent locomotion, cultural and personality influences, and environmental sensing.

The course speakers represent four of the major research ``streams'' in real-time Virtual Humans: Badler and Metaxas with the Jack software and its extensions and derivatives; Perlin and Goldberg with the IMPROV software that was featured live at SIGGRAPH '95 and '96 and at our course in '97; Magnenat-Thalmann illustrating the top Virtual Human laboratory in Europe, and Bruderlin who contributed to the LifeForms software and now offers a Hollywood view from Sony Pictures Imageworks. These groups collectively encompass the major issues and proposed solutions to real-time Virtual Human animation. This course will bring them together to share their approaches and experiences as well as their visions for the future.

1. (Badler) Virtual Humans
-- Behaviors for bodies: inverse kinematics, balance, climbing, collision avoidance, and optimal control
-- Planning, scripting, and acting
-- Attention and awareness
-- Multi-agent task simulation
-- Language-based animation control

2. (Bruderlin) Interactive, Procedural Motion Control
-- Walking and Running
-- Motion warping and signal processing

3. (Perlin and Goldberg) Interacting with Virtual Actors
-- Motion texture synthesis
-- Behavior engines
-- Actor scripting
-- Actor personality and reactivity
-- Live performance animation

4. (Magnenat-Thalmann) Networked collaborative Virtual Environments
-- Clothing virtual humans
-- Virtual Life Network
-- MPEG-4 SNHC sound, face, and body animation
-- CyberDance
-- CyberTennis

5. (Metaxas) Physics-Based Models and Motions
-- Dynamics and constraint definition for rigid and flexible objects
-- Finite element techniques
-- Control theory techniques for low level behaviors such as: target tracking, trajectory following, walking under different loads
-- Transitions between walking and running
-- Collision handling techniques for articulated objects
-- Video-based shape and motion capture for face and articulated segments

6. Panel (All speakers): The Future of Virtual Human Animation

This course is intended to demonstrate the state-of-the-art in motion synthesis for virtual humans. It covers the essential elements for interactive, real-time, synthesized, and networked animation via behavior, physics, signal processing, scripting, and reactive techniques.

This course will describe the state-of-the-art in interactive, real-time, and networked human motion synthesis and actor behaviors. It will demonstrate real-time human animation generated by techniques such as inverse kinematics, dynamics, and video motion capture, and by higher-level approaches such as behaviorally-scripted agents, personality profiles, and interpersonal and environmental reactivity. Various applications illustrate real-time synthetic humans in virtual prototyping, team tasks, synthetic actors, human-like avatars, language-based interfaces, dance, tennis, and video motion capture.

This course is intended for animators, managers, software developers, students and scientists interested in acquiring or synthesizing human-like character motions, especially in real-time.

Some mathematics background will be useful in understanding the more advanced control techniques, but all will be amply illustrated with working systems.

Official Schedule:

(don't blame me for the breaks!)

8:30 - 8:40 Badler (Welcome)
8:40 - 9:45 Badler (65 min. including questions)
9:45 - 10:00 Bruderlin (15 min.)
10:00 - 10:15 (break) (15 min.)
10:15 - 11:00 Bruderlin (45 min. including questions)
11:00 - 12:00 Perlin/Goldberg (60 min. including questions)
12:00 - 1:30 (lunch)
1:30 - 1:45 Perlin/Goldberg (15 min. including questions)
1:45 - 3:00 Magnenat-Thalmann (75 min. including questions)
3:00 - 3:15 (break) (15 min.)
3:15 - 4:30 Metaxas (75 min. including questions)
4:30 - 5:00 Panel (all) (30min.)