User:Mattoverby/GSoC2020/mattoverby proposal

Volumetric Soft Body Simulation

Name & Contact

  • Matthew Overby
  • Blender ID: mattoverby


Elastic deformation is an essential component of animation pipelines. Volumetric soft body simulation can be used to capture the complex details of muscle and skin, e.g. volume preservation under collision/compression. In visual effects, such methods must also be robust to extreme deformation (no numerical blow-ups) and fast enough for interactive manipulation. Moreover, many of the underlying material models are nonlinear and computationally expensive, such as hyper-elastic models for realistic tissue or artistically-defined deformation. Toward these challenges I propose the implementation of a volumetric soft body solver in Blender, that is not only rapid and robust, but is general to hyper-elastic materials. The solver will also be capable of robustly resolving self-collisions and other constraints. This would greatly improve the use of physics in Blender for animation pipelines.


Blender does not have a volumetric soft body solver. The current placeholder treats soft bodies as a stiff thin shell. This approximation greatly limits the visual quality and realism of the simulation in a number of ways: no volume preservation, poor momentum conservation, inaccurate collision response, and tedious workarounds (if any exist) to combat the aforementioned problems. By implementing a true soft body solver in Blender, it will improve the usage of physical simulation in production environments for deformation tasks. In the future, the solver may serve artists and animators for a variety of functions. For instance, modifications of the proposed solver (ADMM-PD, described in project details) has already found use in the visual effects industry for geometric modeling [1], skin-slide deformation [2], and physical simulation of layered muscle/tissue [3].


I will implement an elastic solver in the Blender code base. This includes:

  • The solver itself
  • Collision detection, either a custom implementation or adaptation of other Blender code
  • Several important (and personal favorite) material models: As-Rigid-as-Possible (ARAP) [4], some Neo-Hookean-ish energy, and Xu-Splines [5]
  • End-user documentation, including information about adjustable parameters and how they impact the visual result of the simulation

A sophisticated user interface for character rigs is not the focus of this project. The goal is the implementation of an optimized soft body solver that is well suited for muscle/skin simulation in character modeling and animation.

Project Details

The solver will be based on ADMM-PD [6]. In short, the method operates by minimizing the equations of motion cast as an optimization problem. It does so through operator splitting, in which momentum and elastic energies are solved separately in an alternating fashion over a number of iterations. The approach is robust to large time steps and extreme deformation, and is accelerated through precomputation and parallelization.

One minor challenge is that Blender does not have direct support for tetrahedral meshes. To handle this, I will use an embedding tetrahedral lattice that is generated as needed. In this approach, the surface vertices are embedded barycentrically within the invisible lattice, and move when the lattice deforms. This has some noteworthy benefits, particularly that the degrees of freedom can be controlled directly (to balance visual fidelity with run time) and the embedded mesh does not need to be watertight or even manifold, which greatly increases usability. Adding tet-mesh input/output or tet-meshing in the future is possible, as the solver (excepting collision detection) is agnostic to what type of tet-mesh is used.

Project Schedule

  • May 18 - June 15: Implementation of the optimized bare-bones soft body solver in Blender. This includes the integrator, rudimentary lattice generation, and ARAP energy.
  • June 16 - July 13: Improve constraint generation and solver. Add support for self-collision and user manipulation.
  • July 14 - August 10: Add additional energy models. Finish any uncompleted tasks from previous milestones.
  • August 10 - August 17: Final write up of documentation. Polish code for readability, comments, and cleanliness.


I am a PhD candidate in Computer Science at the University of Minnesota. I research methods for physics-based animation with a focus on elastic deformation, and I am the author of ADMM-PD [6]. Some of my previous work is hosted on my GitHub. Other projects (e.g., microclimate simulation) and work experience can be found on my home page.

Over the years I have benefited greatly from open source software and dev libraries. Most of the work I’ve done could not have been completed without them. It’s been a desire of mine to give back something useful for a long time now, and GSoC presents a great opportunity. Blender needs a soft body solver that is well suited for character/creature defo, and I have been looking for an excuse to port our solver to Blender.


[1] David Minor. 2018. Making space for cloth simulations using energy minimization. In ACM SIGGRAPH 2018 Talks (SIGGRAPH ’18). Article 41, 1–2. DOI:

[2] Jun Saito and Simon Yuen. 2017. Efficient and robust skin slide simulation. In Proceedings of the ACM SIGGRAPH Digital Production Symposium (DigiPro ’17). Article 10, 1–6. DOI:

[3] David Minor and David Corral. 2018. Smeat: ADMM based tools for character deformation. In SIGGRAPH Asia 2018 Technical Briefs (SA ’18). Article 2, 1–4. DOI:

[4] Olga Sorkine and Marc Alexa. 2007. As-rigid-as-possible surface modeling. In Proceedings of the fifth Eurographics symposium on Geometry processing (SGP ’07). Eurographics Association, Goslar, DEU, 109–116.

[5] Hongyi Xu, Funshing Sin, Yufeng Zhu, and Jernej Barbič. 2015. Nonlinear material design using principal stretches. ACM Trans. Graph. 34, 4, Article 75 (July 2015), 11 pages. DOI:

[6] M. Overby, G. E. Brown, J. Li and R. Narain, "ADMM ⊆ Projective Dynamics: Fast Simulation of Hyperelastic Models with Dynamic Constraints," in IEEE TVCG, vol. 23, no. 10, pp. 2222-2234, 1 Oct. 2017. DOI: