Real-time gesture animation and collision avoidance

Procedurally generating “natural-appearing” arm gestures and task-dependent movements for a computer graphics character must consider human biomechanical features and limitations. In particular, collision avoidance is often ignored for communicative gestures and a “robotic” look is often generated. We generate natural human motion by using joint strength models that avoid costly optimization approaches or motion capture retargeting. Joints do not all move simultaneously during reaches, so empirical joint recruitment data is used rather than motion modification approaches that manipulate all body joints, no matter what the target. The real-time computation allows the dynamic adjustment of the reach target; this has enabled a follow-on application (funded by an NSF SGER) to real-time American Sign Language generation with realistic arm motions well beyond the current state of “signing avatars”.

Primary funding: Lockheed-Martin “Human Modeling Testbed” ( N. Badler , PI), and N SF American Sign Language Natural Language Generation and Machine Translation (N. Badler and M. Marcus, co-PIs).

•  Y. Liu and N. Badler. “Real-time reach planning for animated characters using hardware acceleration.” Computer Animation and Social Agents, IEEE Computer Society, New Brunswick, NJ, May 2003, pp. 86-93.

•  L. Zhao, Y. Liu and N. Badler. “Applying empirical data on upper torso movement to real-time collision-free reach tasks.” SAE Digital Human Modeling Conference, Iowa City , IA , 2005; published as Paper 2005-01-2685, SAE Trans. J. of Passenger Cars – Mechanical Systems .