tyFlow Forum
physx collision mesh - Printable Version

+- tyFlow Forum (https://forum.tyflow.com)
+-- Forum: tyFlow Discussion (https://forum.tyflow.com/forum-1.html)
+--- Forum: Bugs and Issues (https://forum.tyflow.com/forum-3.html)
+--- Thread: physx collision mesh (/thread-4064.html)



physx collision mesh - nayan - 02-01-2024

this is the first time I have noticed,
if there is no velocity on particle object, then physx collision mesh - hull type "Mesh" which is animated doesnt have any impact on the physx shape particles....this is wierd....I think physx collision mesh doesnt not respect animated mesh transformation as velocity. 
But if I throw the particle on the physx collision mesh object - works.....
Kinematic not working....

The same problem arises if two different events are made with one physx shape particle and other with physx shape mesh as kinematic, again if the particle doesnt have any velocity, kinematic mesh doesnt not collide at all.....

Is this a bug or an issue or this is how physx work?

Thank you
Nayan Bodawala


RE: physx collision mesh - wlfaos2 - 02-05-2024

I'm having the same problem and looking for a solution


RE: physx collision mesh - tyFlow - 02-06-2024

Yes, that's a limitation of the current PhysX implementation in tyFlow...if a collider (in mesh mode) is animated, there is no interpolation happening between frames...the mesh is simply cleared, and then re-added in its new position...so the "movement" of faces is not really movement at all, PhysX colliders (in mesh mode) are just per-frame snapshots, so if you see them moving "through" particles, it's because they're appearing on one side on one frame, then appearing on the other in the next frame...hence why no collision is triggered if the particles themselves are not moving. The same is not true for the other collision modes...they can trigger collisions even for stationary particles.

PhysX has some built-in methods to do interpolation in mesh mode (I haven't tested them, but I imagine they overcome this problem)...however, it's not as simple as implementing that directly...tyFlow allows users to input meshes with changing topology as a PhysX collider, or meshes with unchanging face counts, but face re-ordering, etc...these would all cause big problems for any mesh collider interpolation, require extra tracking, etc. Perhaps I could simply add a checkbox to enable/disable the interpolation function, in case of these other situations where it wouldn't work properly....I will add that to the todo list as something to explore.


RE: physx collision mesh - d4rk3lf - 02-06-2024

This is why I often use, when having animated mesh collider, just another small flow with birth object, voronoi, object bind, PhysX Shape and PhysX Switch (Kinematic), for that.
Worked pretty nicely so far.


RE: physx collision mesh - nayan - 02-07-2024

Hey tyson,
This issue is very limited to the ver 1.1 and above
there was never any problem or issue in ver1.029, everything is working absolutely perfect. Even if particles had zero velocity, physx collision with mesh hull was working beautifully
I have personally checked both ver 1.029 and the latest ver 1.102, I also asked fellow artist - mohammed hejair and walid al zayani and they are also getting the same result what I am getting. So i guess there is some kind of bug

This is not limited to physx collision mesh alone, even if I birth any animated geometry(just a normal animation in x axis), physx shape mesh will also not work on any particle,

Please look into the issue

Thank you
Nayan Bodawala

(02-06-2024, 12:31 AM)tyFlow Wrote: Yes, that's a limitation of the current PhysX implementation in tyFlow...if a collider (in mesh mode) is animated, there is no interpolation happening between frames...the mesh is simply cleared, and then re-added in its new position...so the "movement" of faces is not really movement at all, PhysX colliders (in mesh mode) are just per-frame snapshots, so if you see them moving "through" particles, it's because they're appearing on one side on one frame, then appearing on the other in the next frame...hence why no collision is triggered if the particles themselves are not moving. The same is not true for the other collision modes...they can trigger collisions even for stationary particles.

PhysX has some built-in methods to do interpolation in mesh mode (I haven't tested them, but I imagine they overcome this problem)...however, it's not as simple as implementing that directly...tyFlow allows users to input meshes with changing topology as a PhysX collider, or meshes with unchanging face counts, but face re-ordering, etc...these would all cause big problems for any mesh collider interpolation, require extra tracking, etc. Perhaps I could simply add a checkbox to enable/disable the interpolation function, in case of these other situations where it wouldn't work properly....I will add that to the todo list as something to explore.

Hey tyson,
This issue is very limited to the ver 1.1 and above
there was never any problem or issue in ver1.029, everything is working absolutely perfect. Even if particles had zero velocity, physx collision with mesh hull was working beautifully
I have personally checked both ver 1.029 and the latest ver 1.102, I also asked fellow artist - mohammed hejair and walid al zayani and they are also getting the same result what I am getting. So i guess there is some kind of bug

This is not limited to physx collision mesh alone, even if I birth any animated geometry(just a normal animation in x axis), physx shape mesh will also not work on any particle,

Please look into the issue

Thank you
Nayan Bodawala


RE: physx collision mesh - tyFlow - 02-07-2024

If you have a file where you can reproduce the issue, please send it my way. My explanation above applies to meshes that move over particles completely (so there's no intermediate overlap)....perhaps you have found a legitimate bug.


RE: physx collision mesh - tyFlow - 02-07-2024

Alright - my mistake: I thought the effect being reported was that of collider tunneling, which can happen if PhysX mesh colliders move too fast, since they update in steps as described above.

The problem Nayan was referring to is an actual bug that was introduced in v1.100, which will be fixed in the next build, where no PhysX Collision colliders in mesh mode update properly when moving.


RE: physx collision mesh - tyFlow - 02-07-2024

If you redownload v1.102 this issue should be fixed.


RE: physx collision mesh - nayan - 02-09-2024

(02-07-2024, 11:04 PM)tyFlow Wrote: If you redownload v1.102 this issue should be fixed.

Thank you


RE: physx collision mesh - wlfaos2 - 02-17-2024

(02-07-2024, 11:04 PM)타이플로우 Wrote: v1.102를 다시 다운로드하면 이 문제가 해결됩니다.
I'm writing version 1.104 now and the problem keeps popping up. If you choose mesh among hull modes in the physx shape, the target you want to destroy just passes through


RE: physx collision mesh - tyFlow - 02-17-2024

It says why right in your screenshot - Mesh mode cannot be used on dynamic particles. This has always been the case. Use convex or compound mode.