Thesis of Guillaume Saupin

Vers la simulation interactive réaliste de corps déformables

Virtual environments have quickly invaded our lives in the last few years. Video games are becoming more and more realistic and provide the user with detailed environments. This kind of environment is becoming a substitute for reality as in the game Second Life. In the industry, these environments are used on a daily basis, and for a much longer time. They have become essential in product design. In medicine, medical imaging, largely based on 3D visualization methods and technologies, is used every day, but mainly for diagnostics. Physicians are reluctant to use these tools to help them during operations as they currently do not fully trust them. There is still a lack of confidence in these technologies, thus surgeons prefer to rely on their own judgement. This reluctance is understandable since virtual environments provided hitherto only graphic information. The objects behaviour in virtual environments is seldom based on physical grounds. Hence, the medical community's defiance and the industry's restlessness for more realistic and physically based virtual environments. By now, the CPU ever increasing computing performance, the general purpose GPU development conjugated with the Computer Graphics community vitality has allowed more and more realistic simulations. Hence, the last few years have seen the emergence of interactive deformable bodies simulations that have been largely used in many applications. Nevertheless, these interactive simulations are subject to many limitations which restrict their use. For instance, the simulated objects cannot collide with their environment, or they collide in an unphysical way. They usually cannot undergo large displacements, and if they are able to, they usually do it in an unrealistic way or using only a small set of deformation modes. In our Ph.d thesis, which is summed up here, we try to find ways to get over these limitations. Our goal was to develop methods to interactively simulate deformable bodies undergoing large displacements in a physical way. We also wanted to handle contact between the deformable objects and their environment using a contact model which was as realistic as possible. And eventually we wanted to add an haptic feedback in order to improve the user immersion.

Jury

Thesis of the team MINT defended on 26/11/2008