plants wind dynamic pipeline

[mosssi]

Hi guys

I'm working on a field plants dynamic and set dressing project like the attached image;

So I'm confused about the pipeline for simulating the wind effect. I used wire solver and I simulated some of plants individually but for scattering those, I should use rotation for different look and now the wind direction won't look right. also simulating all the plants together is not an easy and efficient task. So is there a better way? can you please give me a hint?

Thank you

[DASD]

I would look into how this is done in modern game engines.
For one, each plant you scatter needs to have blend-shapes in all four directions. The deformation to those blend-shapes should be driven by parameters representing wind direction and intensity. Then you would create a "wind" in world-space and let it animate those parameters.
- Or something like that.

[mosssi]

I would look into how this is done in modern game engines.

For one, each plant you scatter needs to have blend-shapes in all four directions. The deformation to those blend-shapes should be driven by parameters representing wind direction and intensity. Then you would create a "wind" in world-space and let it animate those parameters.

- Or something like that.

 

Thanks Dimitri

I think this is what I want to do so far but blendshape is new and I think that should be really interesting in some situations.

[Farmfield]

You can get pretty far using point displacement in XZ with the noisefield displaced in Z + a bit in the wind direction. 

[mosssi]

You can get pretty far using point displacement in XZ with the noisefield displaced in Z + a bit in the wind direction.

Do you mean using a vop sop instead of dops and make turbulence movement instead of wind? Can you explain a little bit more. Thanks

[Farmfield]

Yeah, pretty much. And it's not overly complicated to setup, adding some noise to the point position in XZ and then you use a ramp to control the strength of the effect in Y as you need to have 0 strength at the bottom and full at the top of the object, it'll slide around on the surface.

 

And this won't work for everything, but for environment setups where you just want to apply some dynamic feel, it's a computationally cheap way to get that.

[mosssi]

Yeah, pretty much. And it's not overly complicated to setup, adding some noise to the point position in XZ and then you use a ramp to control the strength of the effect in Y as you need to have 0 strength at the bottom and full at the top of the object, it'll slide around on the surface.

 

And this won't work for everything, but for environment setups where you just want to apply some dynamic feel, it's a computationally cheap way to get that.

 

That's what I did for my far plants which are cross section planes but for mid and near plants I think it's not possible. I have foreground plants that needs to be simulated. Thanks anyway

[Farmfield]

For hero plants, no, you need something more accurate. I'd try simming those in two stages, perhaps combining FEM and the grain solver. First do a FEM sim for the stems, bake that animation, then a separate sim on top for the leaves. For the leaves you'd just play around and see what looks best, FEM or grains. Grains can achieve really good cloth like behavior, but FEM is likely computationally cheaper - and FEM might also be easier for simming that semi-stiffness leaves have.

Edited January 5, 2016 by Farmfield

[mosssi]

Thank you Farmfield

Now I have some idea to work on. I didnt think about FEM for this before, I know my wire sims are the best result for per plants dynamic but if FEM can handle a field computation reasonably, I might go with that. I will post my experience in some days. 

[Farmfield]

No prob's & good luck. Looking forward to see what you end up with.

 

And on a sidenote, it seems many miss FEM even though the Houdini dev's definitely have not, there's been a huge optimization of FEM in Houdini 15, I believe it's something like 10x faster than in H14.

[michael]

an idea that i had a while back was to build plants/trees/etc as crowd agents...you'd get all the distribution and variation features from the crowd system and adding ragdoll to them would allow for some pretty nice automatic deformations from collisions etc...

[mosssi]

No prob's & good luck. Looking forward to see what you end up with.

 

And on a sidenote, it seems many miss FEM even though the Houdini dev's definitely have not, there's been a huge optimization of FEM in Houdini 15, I believe it's something like 10x faster than in H14.

 

I hope so, FEM was not stable enough in my first experiences in H14 last year.

 

an idea that i had a while back was to build plants/trees/etc as crowd agents...you'd get all the distribution and variation features from the crowd system and adding ragdoll to them would allow for some pretty nice automatic deformations from collisions etc...

 

This is doable Michael. That's the method that Weta used for Avatar jungle. they exactly used Massive for scattering and dynamic distribution but the idea was a little out of my league because I don't have crowd experience yet. but now that you mentioned Houdini crowd, can you compare briefly Houdini crowd and other crowd software like Golaem or Massive. I just want to know if its worthy to learn.

Thank you