The originality in our approaches for virtual environment generation and management is two folds. First we include constraints, properties and knowledge in the design, generation and management steps. Then we include learning and autonomous behavior into simulations.
Knowledge-based scene modeling and animation. The creation of huge and complex virtual environments using common “by objects” approaches is a very long and tremendous task. A way to overcome this is to rely on a priori or retrieved scene knowledge (semantics, special/temporal relationships, properties, specific rules, constrains). To generate virtual worlds, we study procedural generation and declarative modeling that are effective to construct the layout, build objects and generate the necessary context to provide with animations. This provides the outputs for visualization and animation systems . To populate virtual worlds, two approaches have been initiated. The first is merging procedural, dynamics and behavioral approaches for animating virtual characters. The second is linking modeling and animation through semantics and knowledge. The DRAMA project provided us resources and a context to successfully validate our approaches .
Knowledge-based scene reconstruction. An important challenge when reconstructing virtual worlds from objects or scenes acquired using 3D scanners that provide huge unstructured point clouds is knowledge recovery. We process geometrical and structural morphology feature extraction from objects surfaces. These features are used to build average reference models, and also for 3D shape classification for objects taxonomy and classification . Our goal is to propose a landmark-free method and a set of data analysis tools fitted to this approach. We also reconstruct large scenes by merging point clouds or face sets of heterogeneous density issued from different acquisitions . When partial/coarse CAD models are available, it is possible to use them for segmenting the points/faces, closing holes and filtering data, and also to use point clouds to add accuracy in a coarse CAD model . Applications are developed for Virtual Heritage and for large telescope infrastructure reconstruction.
Networked virtual/augmented environments. After years working on communication architecture optimization for networked virtual environments , we focused on the study of specific distribution schemes. The goal is to allow users to share a common feeling of presence  in virtual and/or augmented reality environments in an asymmetric relation (using heterogeneous computational environments). We also investigate user “telepresence” within a simulated real environment, build on our previous work on interoperability between heterogeneous distributed simulations. We experimented our solutions in industrial maintenance contexts, for instance with Airbus in the @MOST project (linked to the AIRSYS Laboratory).
Serious Games. Federating several skills in the team, we develop a transversal axis on Serious Game design and development. First, we made a morphologic analysis of serious games to identify basic components. This led us to establish a gameplay component cartography for serious game classification  and to define theoretical foundations for Serious Games design. We have confronted and validated our theory by implementing the Prog&Play software environment. It is a serious game dedicated to strengthen student skills in programming fundamentals . Our activities allowed us to joint the European Network of Excellence GaLA dedicated to Serious Games.
We will develop methods to link generating virtual environment and designing usage and interaction capabilities in a systemic approach fitted to extend 3D environment design. We want to reinforce what is our particularity in the scientific landscape : to use in synergy procedural, metaheuristic and bio inspired approaches when modeling scripted 3D environments. The main scientific challenge we want to tackle in the next five years is the generation in a unified process of enriched 3D environments, in the scope of generating virtual/augmented, populated and interactive 3D environments. We will particularly focus on mechanics to support 3D environment generation and management. The goal is to build software architectures for enriched vivid Virtual Environment in which we will create and manage links between environment component (objects, 3D worlds, virtual characters…) properties and features such as shape, semantics, aspect, animation (motion, script, scenario), behavior, and interaction capabilities.
Virtual characters. Coupling artificial intelligence and character animation allows automatically generating behaviors of virtual entities and associated motions. Behavioral simulation and planning methods can provide more and more complex strategies, tackling manipulation of several interactive virtual objects. Thanks to several skills available in our group (environment modeling, virtual human animation and behavioral simulation) we will generate automatically character animation for interactive virtual object manipulations. This will be based on semantic environment including knowledge on object uses and functionalities.
Interactive scripted adaptive virtual worlds. We will study generating scripting, simulation and interaction processes for applications embedded in virtual environments such as simulations or serious games. Indeed, generating scenario, script, autonomous behavior and interaction allows defining dynamic and interactive content of the virtual world. The upper approach will be used as high level of detail animation. Scenario, scripts and behaviors, but also properties and features would be self-adaptive and evolutionary in order to tackle changes in the environment or with respect to users. This will be achieved thanks to artificial life and bio inspired methods developed in the ALIFE part of the team.
|The 3D Virtual Operating Room (3DVOR) project is an immersive collaborative learning game which targets various defects in healthcare inside the operating room due to a communication failure between the surgeons, the nursing staff, the anaesthetics, the patient and other caregivers. The game allows a team of professionals to re-enact in an immersive and realistic environment numerous scenarios inspired by recurrent routine-induced mistakes or uncommon acute crisis situations, and evaluates their response against prevention and risk management procedures (institution procedures, patient check-list, etc.).|
|3DVOR game interface. Nurse’s point of view.|
|IEEE Conference on Computational Intelligence and Games (GIG 2014), Dortmund, Germany, 2014.|
|3DVOR is supported by the 12th innovation cluster French funding scheme “Fonds Unique Interministériel” (FUI). 3DVOR is a collaborative research project between KTM Advance, Novamotion, the University of Toulouse, Toulouse Hospital and the University JF Champollion.|
|Prog&Play is a library designed for Real Time Strategy games (RTS). It enables player to program easily and interactively units of the video game. Currently, Prog&Play is integrated into Kernel Panic, a multiplayer RTS. Prog&Play and Kernel Panic enable to design serious games for programming practice. It is used for teaching computerg graphics in several institutions.|
|Prog & Play Test field|
|Computer Graphics Forum, Vol. 30 N. 1, p. 61-73, march 2011.|
|The fossil specimens investigated so far represent three relatively well-preserved Australopithecus africanus endocasts : the Taung child and the specimens Sts 5 and Sts 60 from Sterkfontein Member 4. Virtual reconstructions were obtained through high-resolution microtomography for Taung (digital@rchive), medical scanner for Sts 5 and surface scanning for Sts 60.
The comparative modern sample used to assess the sulcal and petalial patterns includes 2 extent humans, EH1 and EH2, respectively an adult and a juvenile individual, 1specimen of Pan troglodytes (Pt) and1 specimen of Pan paniscus (Pp). In the perspective of assessing the morphostructural organization, we analyzed a larger sample composed of 30 adult endocrania of EH (n=10), Pt (n=10) and Pp (n=10).
Petalia We defined a sagittal plane for symmetry as the best-fit plane passing through the landmarks placed along the interhemispheric ridge. Each hemisphere was mirrored along this plane. For each vertex of the original surface, we determined the closest point belonging to the corresponding opposite surface.
Sulcal pattern Based on a method for detecting crest lines on surfaces, the sulcal lines were automatically assessed and manually corrected by removing non-anatomical structures.
Deformation Concerning the deformation based methodology, a 3D endocrania atlas was estimated for each modern species. Given a set of endocrania from the same species, the atlas is seen as a mean template and the deformation from this template to each specimen, therefore integrating the population-related variability. We highlighted the structural differences of the morphostructural organization between extent species and the fossil specimens Sts 5 and Sts 60 by computing the morphological deformation between the common templates of the modern and the two latter.
|Morphological deformations from the mean template (at the top) of extant humans, chimpanzees and bonobos to the fossil specimens Sts 5 and Sts 60. Deformations have been rendered by a color-scale and projected on the original surfaces for more precise visualization.|
|American Journal of Physical Anthropology, Vol. 153, Issue S58, p. 110, 2014.|