Physical-based
animation of deformable objects has gained considerable attention in the last
two decades among the computer graphics community. The need for plausible real
time animations has generated a number of approaches, most of them focused on
the simulation of cloth and facial expressions. Currently, in order to achieve
interactive frame rates, explicit integration and mass-spring models must
be used. FEM-based models and implicit integration results in the need for
solving large linear systems, which makes them unfeasible for interactive
applications.
However, using
explicit methods leads to instability problems when solving stiff equations
for large time steps: the time step must decreased rapidly with stiffer
equations. The interactive simulation of stiff deformable objects is thus
an unsolved problem. Therefore, since interactive frame rates must be achieved in order to
present useful feedback to the user, this problem is well suited for the
application of the high programmability of current graphics hardware.
In the last years, GPUs have
become high performance SIMD processing units. The new features of last
graphics engines approaches them to more general purpose processing units which
have been utilized recently to perform tasks not directly related to the
original purpose they were built for (see GPGPU). In this project, we
exploit these new features provided to develop a GPU-based simulation system for
deformable tetrahedral meshes. Our simulator includes implicit and explicit
integration methods for solving the equation describing the dynamics of the
physical system. To increase the range of applications to surgical training and
pre-operative planning, we coupled our simulation system with a GPU-based
raycaster for tetrahedral meshes, modified to support deformable meshes. This
will provide the user with insights of the inner structures of the deformed
model, which is a key issue in medical applications.
Technical details about the
implementation can be found in an accompanying publication to appear in the Simulation Theory and Practice Journal.
Results
Some examples of
the results obtained are shown below. Promising frame ratios were achieved with
both explicit and implicit methods. Further numerical methods are
to be implemented and included in the simulator.
[TE05] E. Tejada
and T. Ertl. Large Steps in GPU-based Deformable Bodies
Simulation, Simulation Theory and Practice. Special
Issue on Special Issue on Programmable Graphics Hardware. To appear . 2005
