Nicolas PRONOST

PhD - Assistant Professor

Games and Virtual Worlds research group
Utrecht University
The Netherlands

I am currently assistant professor in the Games and Virtual Worlds research group at Utrecht University (The Netherlands). I studied Computer Science at the University of Rennes 1 (France), where I obtained my master degree in 2003. Then, I worked within the SIAMES Project of the INRIA/IRISA Laboratory (Rennes, France) obtaining my PhD degree in Computer Science from the University of Rennes 1 in December 2006. I have worked there as teaching assistant until September 2007. And at that time, I went, thanks to an INRIA associated team, to the State Key Lab of CAD & CG at Zhejiang University (China) as a postdoctoral researcher during 9 months. Then, I joined the Virtual Reality Laboratory (VRLab, EPFL - Switzerland) in a Marie Curie FP6 Research Project (3D Anatomical Human) in September 2008. Finally, in september 2010 I became a member of the Games and Virtual Worlds group at Utrecht University (The Netherlands). My main research topics are :

Biomechanical simulation and character animation [Games and Virtual Worlds - Utrecht University] During my past experiences I came across a number of techniques to analyze human motions and to animate virtual characters. Many common tools such as inverse kinematics and physics-based algorithms participate in filling the gap between these disciplines. Then I believe the next step involves more realistic character-specific simulations. For instance, real-time precise calculations of the alterations involved in the underlying models (e.g. musculoskeletal injuries or physical muscle fatigue) highly influence the naturalness of the virtual characters. My research focuses on the usage of biomechanical simulations (physics-based and Finite Element simulations) within a real-time virtual environment (animation engine for games and interactive applications).

Subject-specific musculoskeletal simulations [EPFL - VRLAB] The common way of creating musculoskeletal models is to use and scale a generic musculoskeletal model based on data derived from anatomical and biomechanical studies of cadaveric specimens. This scaling has been reported to introduce several errors because it does not always account for subject specific anatomical differences. I use a new semi-automatic workflow for creating subject-specific musculoskeletal models from Magnetic Resonance Imaging (MRI) datasets and motion capture data. An anatomical model is reconstructed by using a model-based automatic segmentation approach, where muscles, tendons, bones and corresponding attachments are identified. This anatomical model coupled with motion capture data, joint kinematics information and muscle-tendons actuators is finally used to create a subject-specific musculoskeletal model.

Subject-specific soft-tissue simulations [EPFL - VRLAB] I propose a numerical solution, based on the Finite Element (FE) method, able to estimate muscles deformations during contraction. Organized around a finite element solver and a volumetric environment, this solution is made of all the modeling and simulation processes from the discretization of the studied domain to the visualization of the results. The materials and properties of the different parts of the FE model are also defined such as the hyperelasticity, a contention model based on inter-meshes neighboring nodes pairing, and the estimation of the nodal forces based on the subject-specific muscular forces and action lines.

Heterogeneous motion database [Zhejiang University - State Key Lab of CAD & CG] This work deals with the real-time simulation, in an immersive environment, of a virtual character efficiently reacting with real users. The retargetting and the solving of spacetime constraints are used to adapt a motion which best fit with the current situation. This motion is selected by a data-driven retrieval algorithm among an average-size database filled-in with motions from several actors. This work is illustrated with an example of interactive kung-fu fighting with a virtual character.

Efficient and intuitive selection of motion [Zhejiang University - State Key Lab of CAD & CG] This project deals with a new approach for efficiently selecting a motion into a database. This time the database is huge and the constraints are given by sensors placed over the user's body. We propose an intuitive approach of retrieval in which the system seeks the query motion using the real movement performed by the user.

Kinematical and dimensional adaptation [IRISA - SIAMES] The method of kinematical adaptation of virtual human motions that I proposed, comprises a retargetting method and an algorithm of interpolation. The retargetting allows the production of plausible motions according to the morphological data of the character. While the interpolation selects and combines several motions from a database according to the most influent parameters on each anatomical axis.

Inverse dynamics evaluation [IRISA - SIAMES] I propose a method of validation of the dynamics of such synthesized motion studying the modifications of their physical properties. I compute, thanks to the inverse dynamics principle, the resulting torques and I compare them to the literature and to experimental data from force-plates measurements. This study separates the influence coming from the parameters of the retargetting and the parameters of the kinematical interpolation.

Normalized forward dynamics synthesis [IRISA - SIAMES] Finally, I propose to use the torques to compute new motions thanks to a forward dynamics simulation. I define a normalization of the torques from the physical properties of the simulated virtual human. Then, I discuss the use of such normalized physical data by a study on a small set of locomotions from the database.