Current computer-aided design (CAD) software is still not able to faithfully reproduce in real-time the mechanical properties of ordinary, everyday deformable objects, such as folded thick clothes, pillows with inner bodies, virtual flesh, etc. This is due to the inherent complexities of these objects which are composed by heterogeneous, nonlinear, anisotropic materials. Deformation of soft bodies is thus achieved manually by skilled designers, employing long computational times and hardware/software set-ups which are costly both in terms of money and time.
In this project, I propose to investigate and develop a 3D animation system able to mimic the dynamics of soft deformable bodies in real-time, achieving similar accuracy of off-line, realistic methods. This will be obtained by exploiting the massively parallel computing capabilities of the modern graphics hardware (i.e., GPUs). Differently from prior work, I plan to use approximated, iterative solvers to model and simulate a wide range of different deformable materials (fabrics, human skin, etc.), and handling the elastic and plastic collisions between the deformable objects and the virtual environment. Despite the high convergence speed offered by this class of solvers, their parallelization is still a less studied problem, even though it may unlock unprecedented performances and lead to new insights in the field of parallel numerical computation and computer graphics.
I have already built a prototype of a similar animation system and the underlying theory has been published in top international research journals and conferences (Eurographics, Computer Graphics Forum, Journal of Graphics Tools, etc.). The obtained performance speed up is approximately 10x compared with the industrial state of the art in this field. However, this prototype does not reproduce the properties of the materials, and it needs to be extended to achieve accuracy, robust collision handling, and breakable and granular (sand) objects, or fluids, as outlined in this project proposal.
I have identified economical interest from Swedish companies involved in home furniture design (Ikea Communication AB), video games (Electronic Arts Ghost AB), digital product development and product information (Semcon Sweden AB). These companies have agreed to provide use cases, input data and validation during the development of this technology. The proposed research has a clear public interest and helps raise the perception of Sweden as a high technology provider. This proposal will facilitate the development of this technology and determine commerciality.
Docent vid Chalmers, Computer Science and Engineering (Chalmers), Interaction Design (Chalmers)
Funding Chalmers participation during 2016–2018