• TOURNIER Maxime
• Born 28th of September, 1984
• French Nationality

• tournier.maxime (at) gmail.com
• 20, rue de la Biscuiterie, #402
  38400 Saint-Martin d'Hères, FRANCE

Ph.D. Computer Animation

Abstract

In this thesis, we propose novel, data-driven representations for human poses, suitable for real-time character animation. In the first part, we exploit Lie group statistical analysis techniques to approximate the pose manifold of a motion capture sequence by a reduced set of geodesic coordinates. We propose an inverse kinematics algorithm using this reduced parametrization to automatically produce poses close to the learning set. We demonstrate the value of the resulting pose model by an application to motion capture data compression, where only a few end-effector trajectories are used to recover a good approximation of the initial data.

In the second part, we extend this approach to the physically-based animation of virtual characters. The reduced parametrization provides generalized coordinates in a Lagrangian formulation of mechanics. We derive an explicit time integrator by approximating existing variational integrators, and propose a damping model based on the Levenberg-Marquardt algorithm. We also describe a geometric, data-driven, angular limit learning algorithm and its associated kinematic constraints.

In the third part, we reach the problem of task-space motion control. By formulating both physical simulation and inverse kinematics as two sequential quadratic programs, we propose a simple pseudo-control algorithm that interpolates between the two metrics. This allows for an intuitive trade-off between uncontrolled simulation and inverse-kinematics. Since this approach makes use of external forces, we propose an alternate formulation by the relaxation of the true, non-convex control problem into a convex quadratic program. These algorithms are evaluated on simple balance and tracking controllers.

Publications

2011	Natural character posing from a large motion database Xiaomao Wu, Maxime Tournier, Lionel Reveret IEEE Computer Graphics and Applications, IEEE, 2011, 31 (3), pp. 69-77
2009	Motion Compression using Principal Geodesic Analysis Maxime Tournier, Xiaomao Wu, Nicolas Courty, Elise Arnaud, Lionel Reveret Proc. of Eurographics. Munich, Germany, March 2009. 3rd Best Paper Award.
2008	Motion Compression using Principal Geodesic Analysis (Poster) Maxime Tournier, Xiaomao Wu, Nicolas Courty, Elise Arnaud, Lionel Reveret Symposium on Computer Animation, Dublin, Ireland, July 2008.
2007	Paramétrisation de données MoCap pour la compression d'animations 3D Maxime Tournier, Lionel Reveret Master Thesis, Institut National Polytechnique de Grenoble, June 2007.

Work Experience

Education

Technical Skills

Personal Interests