The Préviz Project

I am currently working on the Préviz project as a research engineer. This is a research and industrial french project in which the leader company is Technicolor in Rennes. Several companies (Technicolor, Ubisoft, SolidAnim, LoumaSystems and Polymorph) as well as academic patners (INSA of Rennes, ENS Louis Lumière, LIRIS and GISPI-Lab) are involved in this project. The purpose is to design new previs tools for visual effects to help movie directors during the shooting of the scenes. To do so, and also to reduce the costs associated to the post-production, we will propose new solutions to handle and enhance real time previs onset.

Our role as INSA, and my role as a research engineer, will be to design new interactions between the real actors and the virtual world. Thus, the acting of those actors will have a real time effects on the evolution of the virtual world. For instance, if a real actor and a virtual robot have to shake hands, the actor will just have to raise his hand and the robot will automatically place its at the right position. The director will be able to directly see in real time the result of this interaction. Another element is that events in the virtual world will also impact the real world, by enhancing feedbacks concerning the state of the virtual worlds to the actors. Virtual actors, explosions, and so on must seem real for the actors so that their acting can be easier and more convincing.

The CORVETTE and SIFORAS Projects

I am working also on two other projects whose topics are quite similar to each other : the CORVETTE and the SIFORAS projects. Both of these projects are focusing on Virtual Environments for Training. Corvette is focusing on the collaborative aspect between users and autonomous agents while Siforas is a project more industrial concentrating on the reprensentation of the data in a global and unique database that can be used in various usecases.

In Corvette, we are handling various kind of collaboration : acting on the same object at the same time, synchronise and act together toward a common goal, using communication to let the other users what has to be done... Our role as INSA is to focus on collaborative interactions on the same objects but also during the scenario proceedings. We are also interested in having the same interactions for real users and autonomous agents. To do so, we proposed an abstraction of the actor that allows us to collaborate seemlessly with other actors in the training, no matter they are controlled by users or autonomous agents.

In Siforas, we are focusing more on the data representation in order to use a same database to design different trainings on various plateforms, from tablet to immersive rooms. As INSA, we are mainly involved in the project for our expertise in Virtual Environments for Training. We are informing the industrial partners of the many pedagogical tools we designed and of how they could integrate them in their usecases.

My Ph.D Thesis

The aim of behavioral animation is to populate virtual environments of autonomous virtual entities. The autonomy of a virtual character is defined by its capabilities to perceive its environment, to decide and realize actions on its own in order to reach an assigned goal. An entity's behavior is generally defined using different capabilities. One of its major capability concerns its navigation abilities. The navigation task rises many problematics such as "How to find a path between from a start to a goal position?", "How should the entity adapts its posture to avoid collisions with the environment?" or " How does the entity avoid dynamic obstacles?". The number of application fields in this domain is increasing : Video Games, Virtual Reality, Product Lifecycle Management, Architecture,... Considering the interactive aspect of those applications, real time algorithms are needed to handle the path planning and the posture adapation tasks.

My research works mainly focus on path planning and posture adaptation in dynamic environments. I am interested in developping real-time solutions for autonomous entities navigation in complex dynamic environments. In this kind of virtual worlds, configuration can change at any time. For instance, we want to handle digital factories or video games worlds using physic simulations. I am currently focusing on path planning in such environments while avoiding dynamic obstacles but also using every available objects on the trajectory to help entities in their navigation tasks.