Skip to content

← ArchivePaper2005

Robust Quasistatic Finite Elements and Flesh Simulation

Joseph Teran, Eftychios Sifakis, Geoffrey Irving, Ronald Fedkiw

SCAAcademic4 descendantsMuscles

Foundational FEM flesh simulation using quasistatic Newton-Raphson iteration robust to element inversion, enabling character soft-tissue deformation.

Abstract

This paper presents a quasistatic finite element algorithm for robustly simulating deformable flesh attached to a kinematic skeleton. To enable fast conjugate gradient solvers during Newton-Raphson iteration, the method modifies the element stiffness matrices to guarantee positive definiteness even under heavy compression and large boundary condition jumps, by diagonalizing the deformation gradient and clamping negative eigenvalues. Building on invertible finite elements, it smoothly extends elastic forces into the inverted and degenerate regime, removing the artificial time step restrictions usually needed to prevent mesh inversion. A penalty based, level set driven strategy is introduced for handling collision and self-collision of deformable tetrahedral bodies, demonstrated on flesh and muscle of the upper torso derived from the visible human data set.

How to read this

Category
Method: robust quasistatic FEM for flesh simulation
Contributions
  • A quasistatic finite element algorithm for simulating deformable flesh attached to a kinematic skeleton
  • Modifies element stiffness matrices (diagonalizing the deformation gradient and clamping negative eigenvalues) to guarantee positive definiteness for fast conjugate-gradient solves, even under heavy compression
  • Extends invertible finite elements into the degenerate regime and adds a penalty-based, level-set-driven strategy for collision and self-collision of tetrahedral bodies
Context
Builds on invertible finite elements and on the authors' muscle work (Creating and Simulating Skeletal Muscle from the Visible Human Data Set), targeting robust soft-tissue deformation for characters.Builds on: Creating and Simulating Skeletal Muscle from the Visible Human Data Set
Correctness
Demonstrated on flesh and muscle of the upper torso from the Visible Human data set; the quasistatic assumption omits true inertial dynamics, so it targets pose-dependent deformation rather than full dynamic response, and collision handling is penalty-based (parameter sensitive).
Clarity
Technically dense; a first pass conveys the robustness goals, but the eigenvalue clamping and invertible-element handling require a careful second or third pass.
How to read it
First pass for the robustness contributions (positive-definiteness fix, inversion handling); plan a slow second/third pass on the linearization and collision sections if you implement an FEM flesh solver.

Builds on

Built upon by

Related work

Keywords

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →