Realistic motion learning and real-time synthesis in virtual reality
Sébastien Quirion
Denis Laurendeau (Supervisor)
Mario Marchand (Co-supervisor)
Problem: Our project, entitled Waza, strives to develop a simple and general approach for learning and synthesizing human-like movements using a physics simulation. One of our particular interests involves locomotion movements.
Motivation: This project is motivated by the photorealism which is sought by video game and animation film producers. Animation using physics based approaches has gradually been designated as a simple and efficient solution to this problem, as witnessed by recent university research and the new commercial products now available. Our project will contribute significantly to the improvements offered by these approaches to animation.
Approach: Due to the interactive nature of the applications, we propose an approach based on forward dynamics. We thus first define a simple approach allowing the development, through automatic learning, of a controller acting an a human biomechanical model so as to produce the desired movement. A module will then be designed to allow the real-time synthesis of the learned movements by the controller. This module will be designed so as to ensure the stability of the character and respect the movement despite the variable conditions of the synthesis environment.
Calendar: Fall 2005 - Winter 2009
Support: This research is supported by NSERC
Web reference:
Last modification: Nov 16 2007 10:10AM by squirion


©2002-. Computer Vision and Systems Laboratory. All rights reserved