James Kuffner  

Autonomous Agents for Real-Time Animation


Advances in computing hardware, software, and network technology have enabled a new class of interactive applications involving 3D animated characters to become increasingly feasible. Many such applications require algorithms that allow both autonomous and user-controlled animated human figures to move naturally and realistically in response to task-level commands. This thesis presents a research framework aimed at facilitating the high-level control of animated characters in real-time virtual environments. The framework design is inspired by research in motion planning, control, and sensing for autonomous mobile robots. In particular, the problem of synthesizing motions for animated characters is approached from the standpoint of modelling and controlling a "virtual robot".

Two important classes of task-level motion control are investigated in detail. First, a technique for quickly synthesizing from navigation goals the collision-free motions for animated human figures in dynamic virtual environments is presented. The method combines a fast 2D path planner, a path-following controller, and cyclic motion capture data to generate the underlying animation. The rendering hardware is used to simulate the visual perception of a character, providing a feedback loop to the overall navigation strategy. Second, a method for automatically generating collision-free human arm motions to complete high-level object grasping and manipulation tasks is formulated. Given a target position and orientation in the workspace, a goal configuration for the arm is computed using an inverse kinematics algorithm that attempts to select a collision-free, natural posture. If successful, a randomized path planner is invoked to search the configuration space (C-space) of the arm, modeled as a 7-DOF kinematic chain, for a collision-free path connecting the arm initial configuration to the goal configuration. Results from experiments using these techniques in an interactive application are presented and evaluated. Finally, how this research fits into the larger context of automatic motion synthesis for animated agents is discussed.

Ph.D. Thesis Files

"Autonomous Agents for Real-Time Animation"
James J. Kuffner, Jr.
Dept. of Computer Science
Stanford University
December 1999

Click image for online description,
images, movies, and links

From this page, you can download PDF and compressed postscript versions of my thesis. The thesis is available in both single-sided and double-sided formats. If your printer prints double-sided, I recommend downloading the double-sided format. Individual chapters can be downloaded separately for those with low bandwidth network connections or limited printer memory.

Download Entire Thesis

175 pages single-sided
180 pages double-sided

Compressed Postscript 4.33 MB  single-sided    double-sided
Adobe PDF 2.92 MB          double-sided

Download Individual Chapters File size Single-sided Double-sided
Title, Preface, Abstract, Table of Contents
53 k
Chapter 1 : Introduction
104 k
Chapter 2 : Character Animation Framework
208 k
Chapter 3 : Goal-Directed Navigation
1.03 MB
Chapter 4 : Sensing the Environment
1.04 MB
Chapter 5 : Manipulation Tasks
1.93 MB
Chapter 6 : High-Level Behaviors
239 k
Chapter 7 : Conclusion
47 k
73 k

BibTeX Citation

  author =	 {J.J. Kuffner},
  title =	 {Autonomous Agents for Real-time Animation},
  school =	 {Stanford University},
  department =	 {Dept. of Computer Science},
  address =	 {Stanford, CA},
  month =	 dec,
  year =	 1999,

1997 - 2009 © James Kuffner, Jr.