Simulation of skeletal muscles in real-time with parallel computing in GPU
Export citation
Abstract
The animation of humanoid characters is an ongoing research area, with a special focus within
the entertainment industry. However, the need of detailed human models has extended to areas
such as medicine, and biomechanics, each with its own set of performance and functional re-
quirements. While visual realism is more desirable in the entertainment industry, biomechanical
accuracy and interactivity are most crucial in designing medical applications.
Creating biomechanically accurate human models is a great challenge because it requires a
precise reconstruction of the different structures of the human body, as well as the biological
and physiological functions that control them.
Specifically, the modeling and simulation of
the skeletal muscles have received special attention because they generate movement and help
maintain the poses of the human body.
Most of the approaches that attempt to simulate the skeletal muscles of the body are based on
models that are not ideal due to several reasons: biomechanical models, which are mechanical
simplifications of the actual behavior of the muscles, are used; focus is given to their macroscopic
behavior, leaving behind the mechanics and internal structures of the muscles; simulations are
not processed in real-time, which is not ideal for specific applications, such as Computer Assisted
Surgery, where interactivity is crucial.
In this work, a meshfree model that simulates skeletal muscles considering their functioning and
control based on electrical activity, their structure based on biological tissue, and that computes
in real-time, is presented. Meshfree methods were used because they are able to surpass most
of the limitations that are present in mesh-based methods. The muscular belly was modelled as
a particle-based viscoelastic fluid, which is controlled using the monodomain model and shape
matching. The smoothed particle hydrodynamics (SPH) method was used to solve both the
fluid dynamics and the electrophysiological model. To analyze the accuracy of the method,
a similar model was implemented with the Finite Element Method (FEM). Both FEM and
SPH methods provide similar solutions of the models in terms of pressure and displacement,
with an error of around 0.09, with up to a 10% difference between them.
The model was
tested with simulations of contraction and extension of the long head of the triceps brachii
and the vastus lateralis. The muscle’s geometry was able to return to its original configuration
after being innvervated with a stimulus current, displaying contractions and bulging similar to
that of a real muscle. Through the use of General-purpose computing on graphics processing
units (GPGPU), real-time simulations, with at least 70 frames per second, that offer a viable
alternative to mesh-based models for interactive biological tissue simulations was achieved.