Modelling

Declarative Modelling:
An important part of our research deals with declarative modelling of entities (objects making up in the environment) and complex virtual environments in an interactive or declarative way. The correct layout of these entities in a virtual environment deals with the description of relations between entities and an automatic search of solutions. Relations and properties must be first interpreted in constraints then solved by an appropriated and efficient constraint solver (the layout problem is a NP-problem). We studies different approaches based on CSP (layout of few critical entities) and metaheuristics (layout of a large number of entities.

Intuitive modelling of free-form 3D shapes:
Nowadays, modelling tools as well as interaction metaphors remain unadapted for non-expert users. We then investigate following two main axes: 3D shape prototyping by sketching in order to permit a fast and easy modelling of a coarse representation of a 3D object, and the improvement of existing shape representations such as implicit surfaces and subdivision surfaces in order to ameliorate existing techniques. On the other hand, we also develop new reconstruction and rendering techniques for point-based representations.

Rendering and Visualization

Physically based rendering follows the physical behaviour2 of light as closely as possible in an effort to predict what the final appearance of a design will be. An efficient simulation needs surfaces with accurate physical properties as input data. The Bidirectional Reflectance Distribution Function (BRDF) is a fundamental local property describing the light reflection on a surface, which exerts an influence upon the global illumination of the scene due to interreflections. In these works, our aims are accurate modelisation of surface properties from measurement using seperated wavelet transforms applied to the directional and the spectral component of BRDF and integration of such representation in a global illumination process. We works as well on interactive global illumination of dynamic scenes.

In real-time rendering, we study point-based representations and algorithms. The motivation of our recent efforts is to directly render a large point set, i.e. to reconstruct a continuous image of the point cloud. To complete this task efficiently, we first propose an hardware accelerated surfel rendering based on the EWA filtering that give us very high texture filtering quality. In order to render scenes of high complexity we show how to take advantage of temporal coherency in a very accurate hardware accelerated point selection algorithm. Our method does not need any preprocessing and allow the expensive computations to be peformed only on visible points like deferred shading techniques.

Animation, behavioural simulation, artificial live

We have worked on low level Motion Control Model and script language linking MCM to behavioural simulation. Our animation plat-form used several MCM concurrently or in collaborative way, using priority and weighted blending, the best MCM is chosen according to action to perform. This action is an order either from the user, either from a script, or from the behavioural simulation module. Motion capture is an easy and powerful way to create motions, but results need to be post processed. We started to work on motion editing (warping, retargeting, ...) using space-time constraints. We use a Motion Star Wireless from Ascension and we will get other devices such as a face tracker and an eye tracker.

The need to produce believable motion has led to initiate characters animation works based on motion capture, as described upper. The will is to provide autonomous characters using behavioural simulation including the ability to learn, to evolve, to adapt themselves to environment evolution, to collaborate and to communicate with other characters or humans. Emerging individual or group behaviours are built using artificial life paradigms. The simulation is based on several tools such as genetic algorithms, neural networks, classifier systems, cognitive maps, and multi criteria optimisation.

Virtual reality

More specifically, we work on aiding to efficiently distribute such a virtual environment on network of workstations or clusters according to the current main standard (HLA : DoD High Level Architecture - IEEE Standard number 1516 and Facility for Distributed Simulation Systems standard 1.0 of the Object Management Group). We also work on distributed virtual environments interoperability and on cooperative interaction in virtual worlds. Our current applications include Virtual Prototyping, Driving Simulations and Firefighter training.

Firefighter training is a collaboration with CHI teams in IRIT, one in the Speech Processing area and the other one in the ergonomic and user feedback field. The goal of this second application is to use a distributed virtual environment to train firemen officer to command according to scenarios. Characters are avatars or autonomous agents. Avatars are semi autonomous agent controlled by high level commands scheduling/triggering generated or captured motions.