Cloth doesn't collide with objects, and itself. #43406

New Issue

Kamil Adamski · 2015-01-25T10:26:20+01:00

Kamil Adamski commented

2015-01-25 10:26:20 +01:00

System Information
Operating system and graphics card
Windows 7 Ultimate, GTX 970.

Blender Version
Broken: All new from Buildbot including Gooseberry branch.
Worked: Official from Blender.org

Short description of error

Like in topic. Cloth object don't collide with collision object, and don't collide with itself.

Exact steps for others to reproduce the error
Add plane, subdivide, add to it cloth simulation. Add sphere, add collision and watch.

**System Information** Operating system and graphics card Windows 7 Ultimate, GTX 970. **Blender Version** Broken: All new from Buildbot including Gooseberry branch. Worked: Official from Blender.org **Short description of error** Like in topic. Cloth object don't collide with collision object, and don't collide with itself. **Exact steps for others to reproduce the error** Add plane, subdivide, add to it cloth simulation. Add sphere, add collision and watch.

Kamil Adamski commented

2015-01-25 10:26:20 +01:00

Changed status to: 'Open'

Kamil Adamski commented

2015-01-25 10:26:20 +01:00

Added subscriber: @adrian2608

blender-admin commented

2015-01-25 10:26:20 +01:00

#43527 was marked as duplicate of this issue

blender-admin commented

2015-01-25 10:26:20 +01:00

#43717 was marked as duplicate of this issue

ronan ducluzeau commented

2015-01-25 15:17:18 +01:00

Added subscriber: @zeauro

ronan ducluzeau commented

2015-01-25 15:17:18 +01:00

I confirm the same problem under Ubuntu 14.04 64 bits.

no_cloth.blend

attached.blend works under official 2.73 but it does not for master 2.73.4 hash 3121941.

I confirm the same problem under Ubuntu 14.04 64 bits. [no_cloth.blend](https://archive.blender.org/developer/F139336/no_cloth.blend) attached.blend works under official 2.73 but it does not for master 2.73.4 hash 3121941.

Lukas Tönne self-assigned this 2015-01-26 16:15:14 +01:00

Lukas Tönne commented

2015-01-26 16:15:14 +01:00

My bad, i disabled contact point generation for cloth during work on the hair solver.

Using the same collision algorithm for cloth (surface) and hair (strands) geometry may not be a good idea in the long run. For a simple penalty-force response it may still work ok-ish, but already the contact detection needs to work differently with simple BVH tree intersection tests.

Eventually i'd like to move to Bullet collision detection, using ghost objects so we can handle responses separately in our own solver. Cloth is a pretty straightforward mesh collision shape, but hair requires some more creative solution, e.g. using a compound implicit shape of cylinders and spheres (connected capsules).

The response part could then at least use a 3-stage Coulomb friction model (free - impacting - sliding), which works a lot better for stable resting on colliders. We could then also look into more advanced collision integration in the solver itself, using a solver method that supports unilateral constraints much better than penalty forces. For in-depth info, here is a selection of papers on the subject:

Large Steps in Cloth Simulation (Baraff, Witkin 1998)
A Hybrid Iterative Solver for Robustly Capturing Coulomb Friction in Hair Dynamics (Daviet, Descoubes 2011)
Implicit Contact Handling for Deformable Objects (MA Otaduy, R Tamstorf, D Steinemann 2009)

My bad, i disabled contact point generation for cloth during work on the hair solver. Using the same collision algorithm for cloth (surface) and hair (strands) geometry may not be a good idea in the long run. For a simple penalty-force response it may still work ok-ish, but already the contact detection needs to work differently with simple BVH tree intersection tests. Eventually i'd like to move to Bullet collision detection, using [ghost objects ](http://bulletphysics.org/Bullet/BulletFull/classbtGhostObject.html) so we can handle responses separately in our own solver. Cloth is a pretty straightforward mesh collision shape, but hair requires some more creative solution, e.g. using a compound implicit shape of cylinders and spheres (connected capsules). The response part could then *at least* use a 3-stage Coulomb friction model (free - impacting - sliding), which works a lot better for stable resting on colliders. We could then also look into more advanced collision integration in the solver itself, using a solver method that supports [unilateral constraints ](http://en.wikipedia.org/wiki/Unilateral_contact) much better than penalty forces. For in-depth info, here is a selection of papers on the subject: [Large Steps in Cloth Simulation (Baraff, Witkin 1998) ](http://www.cs.cmu.edu/~baraff/papers/sig98.pdf) [A Hybrid Iterative Solver for Robustly Capturing Coulomb Friction in Hair Dynamics (Daviet, Descoubes 2011) ](https://hal.archives-ouvertes.fr/hal-00647497/document) [Implicit Contact Handling for Deformable Objects (MA Otaduy, R Tamstorf, D Steinemann 2009) ](http://www.gmrv.es/Publications/2009/OTSG09/OTSG09.pdf)

Sergey Sharybin commented

2015-01-27 11:29:24 +01:00

Added subscriber: @Sergey

Zauber Paracelsus commented

2015-01-28 02:50:10 +01:00

Added subscriber: @ZauberParacelsus

Kévin Dietrich commented

2015-02-02 19:41:58 +01:00

Added subscriber: @tischbein3-2

Lapineige commented

2015-02-15 18:19:57 +01:00

Added subscriber: @Lapineige

Kévin Dietrich commented

2015-02-18 15:30:01 +01:00

Closed as duplicate of #43717

Kévin Dietrich closed this issue

2015-02-18 15:30:01 +01:00

Kévin Dietrich commented

2015-02-18 15:30:53 +01:00

Changed status from 'Duplicate' to: 'Open'

Kévin Dietrich reopened this issue

2015-02-18 15:30:53 +01:00

Kévin Dietrich commented

2015-02-18 15:30:53 +01:00

Added subscriber: @kevindietrich

Kévin Dietrich commented

2015-02-18 15:30:53 +01:00

Sorry I screwed up merging duplicates...

Kévin Dietrich commented

2015-02-18 15:31:42 +01:00

Added subscribers: @sindra1961, @LukasTonne

Rajaditya Mukherjee commented

2015-02-22 20:32:38 +01:00

Added subscriber: @rajaditya_m

Rajaditya Mukherjee commented

2015-02-22 20:32:38 +01:00

Hi I am a newbie with who is willing to work on cloth simulation aspect of blender since it is a part of my Phd thesis. Is this bug already fixed ?

Zauber Paracelsus commented

2015-02-22 20:34:30 +01:00

In #43406#290864, @rajaditya_m wrote:
Hi I am a newbie with who is willing to work on cloth simulation aspect of blender since it is a part of my Phd thesis. Is this bug already fixed ?

I suggest using the 2.73a release. The issue with cloth collisions only occurs in the upstream development versions of blender.

> In #43406#290864, @rajaditya_m wrote: > Hi I am a newbie with who is willing to work on cloth simulation aspect of blender since it is a part of my Phd thesis. Is this bug already fixed ? I suggest using the 2.73a release. The issue with cloth collisions only occurs in the upstream development versions of blender.

Rajaditya Mukherjee commented

2015-02-23 01:45:03 +01:00

Yes, I tested it with the 2.72 release edition and it seems there the problem is not there. But it is occuring in the developer build that I got from the git repository. I will look into this. Since this is my first tryst with blender any suggestions will be welcome.

blender-admin commented

2015-03-01 12:33:30 +01:00

This issue was referenced by dfefd36837

This issue was referenced by dfefd3683710a4bf70cf90c874b65a9c118d2c06

Lukas Tönne commented

2015-03-01 12:38:02 +01:00

Changed status from 'Open' to: 'Resolved'

Lukas Tönne closed this issue

2015-03-01 12:38:02 +01:00

Lukas Tönne commented

2015-03-01 12:38:02 +01:00

Closed by commit dfefd36837.

Closed by commit dfefd36837.

perfection cat commented

2015-03-02 02:34:55 +01:00

I seem to recognize collision, but do not function for cloth sim.
I attach a file.
Cube should stand in the former version.
However, bending is ignored in the revised version.
clothtest.blend

[The edition that I confirmed]
blender-2.73-56935e2-win32.zip
version 2.73 (sub 9), branch b'master', commit date b'2015-03-02' b'00:05', hash b'56935e2', b'Release'

I seem to recognize collision, but do not function for cloth sim. I attach a file. Cube should stand in the former version. However, bending is ignored in the revised version. [clothtest.blend](https://archive.blender.org/developer/F146633/clothtest.blend) [The edition that I confirmed] blender-2.73-56935e2-win32.zip version 2.73 (sub 9), branch b'master', commit date b'2015-03-02' b'00:05', hash b'56935e2', b'Release'

perfection cat commented

2015-03-02 02:43:52 +01:00

I think that friction does not function in the same way.

perfection cat commented

2015-03-02 03:29:32 +01:00

I think that some situation is different.
Results are different even if they make the same setting in 2.73a(release edition) and the revision.
The result is the same even if I carry out the result of made blend file in any version.
Made blend files are different.
I made this file in a revised version.
clothtest2.blend
I made this file in 2.73a(release edition).
clothtest2_release.blend
Both setting is the same.

I think that some situation is different. Results are different even if they make the same setting in 2.73a(release edition) and the revision. The result is the same even if I carry out the result of made blend file in any version. Made blend files are different. I made this file in a revised version. [clothtest2.blend](https://archive.blender.org/developer/F146638/clothtest2.blend) I made this file in 2.73a(release edition). [clothtest2_release.blend](https://archive.blender.org/developer/F146639/clothtest2_release.blend) Both setting is the same.

Lukas Tönne commented

2015-03-02 09:59:55 +01:00

Changed status from 'Resolved' to: 'Open'

Lukas Tönne reopened this issue

2015-03-02 09:59:55 +01:00

Lukas Tönne commented

2015-03-02 09:59:55 +01:00

Confirmed, i think i'm not applying positions in quite the same way or so. Will look into it ...