|
The MimeTIC reseach team focus on designing methods for simulating virtual humans (VH) that behave in a realistic manner and act with realistic motions. “Realistic” means that a real human would have behave and moved the same way in the same situation. It involves understanding how real people control their motion and behave in order to design models capable of generating realistic virtual humans.
From the behavioral point of view, VH have to be endowed with decisional capabilities to enable their use for simulation purpose or in interactive applications involving cooperation between real and virtual humans. Consequently, our goal is to propose a unified architecture to model individual and collective behaviors. The decisional architecture needs to include reactive, cognitive, rational and social skills but also needs to manage individual and collective behaviors. This architecture will tackle problems relating to (1) behavior modeling and description and (2) the link between behavioral models (decision), motion control (action) and the environment (perception and action consequences).
In addition to the decisional process, the way the motion is performed must be perceived as realistic by users who have to interact with VH. To this end, VH should satisfy a set of constraints. On the one hand, it has to deal with kinematic and dynamic constraints imposed by the environment. On the other hand, it must be controllable by the behavioral model. As the number of constraints is generally lower than the number of actuators, infinity of solutions exists and selecting one particular solution is a complex problem. Indeed, only a small subset of solutions may be perceived as realistic by a user. The selection process should consequently take biomechanical, neurophysiological and psychological knowledge into account to find one of these realistic solutions. The team will take some of this knowledge into account and will collaborate with experts in psychology when necessary.
All the above problems are of great importance in situations involving natural and realistic interactions. It can be an interaction between the VH and its environment (such as dealing with complex grounds, obstacles, object manipulation…), but also with another VH. It raises new difficulties because the combination of potential solutions increases in a combinatory manner: the unbalance between constraints and actuators increases. When interacting with a user, the main problem is that the VH has to react to unpredictable events that can occur at any time. Moreover, as the VH directly interacts with the user, this latter may be more sensitive to any strange behavior. This enforces the need of providing efficient algorithms that generate believable behaviors and motions.
In this team, we propose to tackle those problems by addressing three challenges that relate to (1) complexity, (2) multi-level control and (3) observation/experimentations.
MimeTIC is based on two laboratories:
|
