Subspace Mixed Finite Elements for Real-Time Heterogeneous Elastodynamics

Ty Trusty, Otman Benchekroun, Eitan Grinspun, Danny M. Kaufman, David I. W. LevinSIGGRAPH Asia 2023 · OpenReview

TL;DR

This paper introduces a novel "Subspace Mixed Finite Element Method" (Subspace MFEM) that enables real-time simulation of complex, heterogeneous elastic materials. By combining a "Skinning Eigenmode" subspace with a material-aware cubature scheme, their method achieves fast convergence and preserves rich rotational motion, overcoming limitations of existing solvers.

What Problem Does This Paper Address?

Simulating the dynamic behavior of elastic objects (elastodynamics) is crucial in many fields, from computer graphics to engineering. While real-time simulations are desirable, existing methods struggle when dealing with objects made of materials with vastly different properties (heterogeneity). These solvers become slow to converge, forcing them to terminate early, which leads to visual artifacts like damping and inaccurate rotations. This paper tackles the challenge of achieving real-time performance for heterogeneous elastodynamics without sacrificing simulation quality.

Methods

The core of the proposed solution is a Subspace Mixed Finite Element Method (Subspace MFEM). Here's a breakdown of the key components:

  • Mixed Finite Element Method (MFEM): The authors build upon a prior MFEM approach that explicitly models both positional degrees of freedom (how points move) and stretch degrees of freedom (how elements deform internally). This formulation is known to handle material heterogeneities well.
  • Skinning Eigenmode Subspace: To accelerate the simulation, they introduce a reduced-order model. Instead of using all possible deformations, they project the problem onto a "Skinning Eigenmode" subspace. This subspace is specifically designed to capture rotational motion effectively, which is a common challenge for traditional subspace methods. Crucially, these skinning weights are "material-aware," meaning they naturally adapt to regions with different material properties.
  • Material-Aware Cubature Approximation: For non-linear elastic energies, which are computationally expensive, they employ a cubature (numerical integration) scheme. This scheme is made "heterogeneity-aware" by clustering the mesh elements based on their skinning weights and then strategically placing cubature points within these clusters. This allows for efficient and accurate approximation of the elastic energy, especially in regions with varying material stiffness.
  • Decoupled Resolution: A significant advantage is that the computational complexity of their method is decoupled from the resolution of the underlying mesh. This means they can use very detailed meshes for visualization while performing computations on a much smaller, reduced-order representation.

Datasets

Not discussed in the paper. The paper uses example geometries (mammoth, crab, sword, candy, pendulum, cantilevered beam, octobot, gatorman) and material properties to demonstrate their method.

Key Insights and Ideas

  • Skinning Eigenmodes for Heterogeneity: The key insight is that "Skinning Eigenmodes," when used as a subspace, are naturally suited for heterogeneous materials because their construction can be made sensitive to material properties. This allows them to capture localized, material-specific deformations better than generic modal subspaces.
  • MFEM + Subspace = Real-time Heterogeneity: Combining the robust formulation of Mixed FEM with the speed of subspace methods is crucial for achieving real-time performance in heterogeneous scenarios. The MFEM aspect ensures good convergence, while the subspace drastically reduces the computational cost.
  • Cubature as a Smart Sampler: The material-aware cubature scheme acts as an intelligent way to approximate complex energies. By clustering based on skinning weights, it ensures that more computational effort is spent on regions that are likely to deform significantly due to material or geometric variations.
  • Decoupling Mesh Resolution from Computation: This is a major practical advantage. It allows for high-fidelity visual representations while keeping the simulation computationally tractable for real-time applications.

Key Results

  • Real-time Performance: The method achieves simulation speeds of up to 120 FPS for complex heterogeneous models (e.g., the mammoth with bone, muscle, and joint materials).
  • Improved Convergence and Reduced Artifacts: Compared to traditional FEM solvers with aggressive timestep termination, their Subspace MFEM shows significantly less damping and more accurate rotational behavior, even with very few solver iterations per timestep. For instance, a crab simulation with only 2 iterations shows much better rotational fidelity than FEM.
  • Speedup: The paper reports an average speedup of over 3 orders of magnitude compared to a full-space MFEM iteration.
  • Material-Awareness Demonstrated: Examples like the candy simulation show how their method accurately captures sharp transitions in material properties, which is difficult for methods using smooth, non-material-aware subspaces.

Limitations and Future Work

  • Global Subspace Artifacts: A limitation of their globally supported subspace is that exciting a local region of the mesh can sometimes lead to noticeable global artifacts (e.g., bending a mammoth's leg causing trunk deformation). They suggest that increasing the subspace size can mitigate this.
  • Overly Energetic Motion: At very low iteration counts, their solver can exhibit "overly energetic" motion, leading to initial jittering artifacts. This is contrasted with FEM's damping artifacts.
  • Future Work:
    • Contact Simulation: Extending their method to handle robust reduced-space contact simulation is a challenging but promising direction.
  • Inverse Design: The solver could be valuable for physics-based inverse design in engineering and biomechanics, where domain heterogeneities are common.
  • Trade-off Exploration: Further understanding the cost/quality trade-off between dense, globally supported subspaces and sparse, locally supported ones could resolve artifacts and improve performance.

Summary updated

Abstract

Real-time elastodynamic solvers are well-suited for the rapid simulation of homogeneous elastic materials, with high-rates generally enabled by aggressive early termination of timestep solves. Unfortunately, the introduction of strong domain heterogeneities can make these solvers slow to converge. Stopping the solve short creates visible damping artifacts and rotational errors. To address these challenges we develop a reduced mixed finite element solver that preserves rich rotational motion, even at low-iteration regimes. Specifically, this solver augments time-step solve optimizations with auxillary stretch degrees of freedom at mesh elements, and maintains consistency with the primary positional degrees of freedoms at mesh nodes via explicit constraints. We make use of a Skinning Eigenmode subspace for our positional degrees of freedom. We accelerate integration of non-linear elastic energies with a cubature approximation, placing stretch degrees of freedom at cubature points. Across a wide range of examples we demonstrate that this subspace is particularly well suited for heterogeneous material simulation. Our resulting method is a subspace mixed finite element method completely decoupled from the resolution of the mesh that is well-suited for real-time simulation of heterogeneous domains.
BibTeX
@inproceedings{trusty2023subspace,
  author = {Ty Trusty and Otman Benchekroun and Eitan Grinspun and Danny M. Kaufman and David I. W. Levin},
  title = {{Subspace Mixed Finite Elements for Real-Time Heterogeneous Elastodynamics}},
  booktitle = {SIGGRAPH Asia},
  year = {2023},
  url = {https://openreview.net/forum?id=ytrpl25w1B},
}
Figure