User:Ishbosamiya/GSoC2021/Proposal

Project Title

Continued Development on Adaptive Cloth Simulator for Blender

Name

Ish Hitesh Bosamiya

Contact

email id: ishbosamiya@gmail.com

blender id/chat: ish_bosamiya

twitter: @ishbosamiya

Synopsis

My GSoC project in 2019, on the Adaptive Cloth Simulator, was an exploration in adding support for parametric on-the-fly remeshing for Blender's Cloth Simulator in order to provide high-resolution-equivalent fidelity at reduced computational cost. During GSoC 2021, I will utilize the insights gained, as well as previously fleshed-out core ideas, to stabilize this feature for production. This will involve introducing a new mesh structure, fixing bugs and regressions, refactoring to integrate well with the latest Blender updates, etc. As a stretch goal, I would improve Blender's cloth caching system to introduce support for Adaptive Remeshing.

Benefits

Artists often need to simulate clothing, but this comes with a variety of technical challenges– mesh topologies need to be manually optimized per scene, a careful balance must be maintained between visual detail and reasonable simulation time by picking the right mesh resolution, etc. all while working on their creative end-goal. The 2019 GSoC project shows that these challenges can be overcome automatically by using Adaptive Cloth Simulation. This year's project would allow for this automation to actually reach artists' hands and help them save crucial time, by providing easy to use parameters that automate the process of finding the optimal mesh for every frame of the simulation.

An additional benefit to this year's project is that it provides the necessary foundation to add Adaptive Cloth Tearing support to Blender.

Deliverables

A production-ready implementation of the Adaptive Cloth Simulator. In particular, I aim to have it ready to put into the review queue to be merged into master. (Note: Actual merging into master is considered out of scope simply because of the duration that such a large review would take.)

Blender's current caching system is unsuitable for situations with changing mesh topology, and thus as a stretch goal, I aim to introduce caching support for systems like the Adaptive Cloth Simulator.

Project Details

The Adaptive Cloth Simulator uses the methods proposed by "Adaptive Anisotropic Remeshing for Cloth Simulation" by Rahul Narain, Armin Samii and James F. O'Brien [1].

GSoC 2019 was an exploratory R&D project to implement the Adaptive Cloth Simulator. During the project, my mentor and I realized that it would not be sufficiently stabilizable to merge into Blender's master branch, due to the large complexity of the project. However, the project helped provide the necessary insights and core implementations that would be needed for a stable version to be built. GSoC 2021 will focus on this stable version of the Adaptive Cloth Simulator, so as to be eventually included in master. In particular, I will rework and refactor the necessary components, in order to better integrate with Blender's updated playback system; and perform the necessary bug-fixes and other miscellaneous improvements to bring the Adaptive Cloth Simulator to practical usability. More details on the GSoC'19 project can be found in my 2019 proposal [2] and final report [3].

Major bugs (in the 2019 implementation) to be fixed:

  • Collapsing of edges doesn't work as expected. In particular, the mesh complexity does not correctly reduce, due to incorrect selection of edges to be collapsed. This leads to a suboptimal simulation. I expect this might require a rewrite of the code that determines "sizing".
  • Some custom-made BMesh operations do not work as expected. It leads to unwanted mesh topology that can sometimes make the simulation unstable. Introducing a new mesh structure can solve this issue and speed up the mesh edit operations. Currently, for the new mesh structure, I am opting for Adjacency and Incidence Framework (AIF) [7] with a few modifications. This might also help with future caching support.

Since GSoC'19, both Blender's dependency graph and its playback system have undergone non-trivial changes that prevent playback of the GSoC'19 implementation of the Adaptive Cloth Simulator. Thus, relevant refactoring will be necessary after understanding these changes.

As a related side note, I could extract the remeshing feature as a separate modifier (instead of being integrated into the cloth simulator). This would allow other modifiers/simulators, such as the soft body simulator, to also take advantage of the remeshing feature. To be discussed with the mentor.

Stretch Goal: Blender's current caching system for the cloth simulator does not support changes in the mesh topology. This is because the system only stores information per vertex. Thus, significant changes to the existing caching system would be necessary. I currently have multiple ideas to approach this– (i) store the complete mesh structure, (ii) store the information per vertex and the changes made to the mesh structure, (iii) run the remeshing step after reading each frame of the vertex-based cache, etc.

Project Schedule

Details Time
Introduce new mesh structure 3 weeks
Fix "collapse edges not running expected number of times" 1 week
Integration with Blender's dependency graph and playback system 2 week
Extensive testing and bug fixes 1-2 week
Completion and finalization of the code 1 week
Reserve time period / stretch goal time 1 week

Schedule Conflicts: None that would impact the above schedule. In the highly unlikely event that a conflict appears, I will coordinate with my mentor.

Bio

I, Ish Bosamiya, am currently doing my B.Tech. in Computer Science and Engineering at PES University, Bangalore, India. As part of GSoC 2019, I worked on Adaptive Remeshing for Cloth Simulation. Since then, independently, I have worked on building a fast and modular cloth simulator from scratch, based upon ideas from multiple research papers (in C++ [4]). I am currently in the process of refactoring and rebuilding it in Rust [5]. In parallel, I have also continued spending time learning about geometry processing. I have made contributions (in my spare time, between academics and other side-projects) to Blender outside of GSoC 2019 [6], and plan to continue doing so after GSoC 2021 too.

References

  1. "Adaptive Anisotropic Remeshing for Cloth Simulation" by Rahul Narain, Armin Samii and James F. O'Brien (SIGGRAPH 2012)
  2. GSoC 2019 Proposal
  3. GSoC 2019 Report
  4. Cloth Simulator in C++
  5. Cloth Simulator in Rust
  6. Blender Contributions
  7. "Adjacency and incidence framework: a data structure for efficient and fast management of multiresolution meshes" by Frutuoso G. M. Silva and Abel J. P. Gomes (GRAPHITE '03)