Jump to content

Macha's Sammelsurium


Macha

Recommended Posts

I've decided to post links to my cabinet of Houdini experiments here.

They are ideas, sketches, WIPs, dead ends, scraps and (hopefully) promising starts and gems.

The first one is a simple random path experiment where a particle travels in integer steps, each step being completely random in either - or + direction of an axis. None of the steps is dependant on the previous, but only on the current, in the manner of a Markov chain.

It creates a discrete fractal that is self-similar on larger scales. If have experiments posted before where the steps are of varying size and directions, effectively this very same process here in the limit and thus creating a type of Brownian motion.

https://vimeo.com/52826663

I went up to 10 million steps. I tried a 100 million step version and while I could calculate and store it (about 20 minutes of computation) I seemed to be just a little too short of memory to get it working inside Houdini (16Gb ram).

The chain of point positions was created outside Houdini with a Python script and then loaded in Houdini.

post-4013-0-98096000-1352094695_thumb.jp

Edited by Macha
  • Like 1
Link to comment
Share on other sites

My second entry is a halo shader! It works on single-voxel volumes, complex volumes, all shapes and forms, multiple lights, micropoly, raytrace and pbr.

I'm an avid halo and skywatcher so I wanted to do an artist controllable shader. I have some more plans with this, so watch this space!

https://vimeo.com/52983161

post-4013-0-35024800-1352432439_thumb.pn

  • Like 2
Link to comment
Share on other sites

My newest entry plays with attractors. Normal ones, strange ones, continuous and discrete dynamical systems!

https://vimeo.com/53512198

To get a surface out of these I randomized some of the parameters, so that what you get is a set of states that represents slightly different systems solved in the same manner.

The colors represent the high or low step numbers.

Python for the ODE solver, vex for the iterating discrete solutions and funky transforms. Garden variety nodes for everything else.

post-4013-0-01364100-1352907920_thumb.pn

  • Like 1
Link to comment
Share on other sites

  • 2 weeks later...

I've worked on a few more things.

Here is a type of discrete fractal:

https://vimeo.com/54713944

Here is an attempt to combine Mathematica with Houdini and calculate shortest paths through networks:

https://vimeo.com/55147619

And finally the Hopf Fibration of a 3-sphere to the familiar 2-sphere:

https://vimeo.com/55241453

post-4013-0-23294300-1355153116_thumb.jp

post-4013-0-44996300-1355153122_thumb.jp

post-4013-0-28559500-1355153162_thumb.jp

post-4013-0-59763500-1355153170_thumb.jp

Link to comment
Share on other sites

Here is a type of discrete fractal:

https://vimeo.com/54713944

"You can transform the replacements as well and get more organic looking shapes but doing this efficiently for more than a few steps is very hard in Houdini."

It should be noted that the generative method isn't very efficient in the first place (regardless if you're using Houdini or not). Rendering the Menger Sponge should in theory be manageable using Kaleidoscopic IFS, which is a type of fractal done using the Distance Estimation method (DE).

http://blog.hvidtfel...ractals-part-i/ (background on the DE method)

http://forums.odforc...brot-primitive/ (Houdini example using DE method)

http://www.fractalfo...cape-time-ifs)/ (menger sponge as DE method)

http://images.math.cnrs.fr/Un-ballon-de-foot-fractal.html (another link with easier to read menger sponge formulas)

http://snipplr.com/view/33781/distance-estimator-for-menger-sponge/ (yet another example for this pic)

Link to comment
Share on other sites

  • 4 months later...

I've long been interested in crowd systems but never had the opportunity to work on one myself. Here's an attempt at identifying several parts of crowd problems.

It's an agent based system with the aim of identifying a target and shooting at it.

The sensory part includes 'sniffing' to identify nearby neighbours and 'seeing' to scan the environment. The brain part tries to process these things and get the neighbour's direction and return avoidance movements. The vision is analyzed for targets. A simple communication system between the agents decides who is going to shoot.

The 'brain' then has to calculate the angle of shooting, given a maximal escape velocity of an imaginary projectile. For many things I like vopsops but when you have to translate equations into nodes it can be a real pain, so that part is packaged into a few vex lines. The trajectory is calculated according to the equations of motion, which is very neat because it automatically takes into account maximal height and does not have to be solved iteratively, meaning you can avoid dops and solve it explicitly at random frames. That makes the solution very farmable indeed. Unfortunately very few differential equations have such solutions, which is why we will need dops for the vast majority of such tasks...

The path of the trajectory until the hitpoint is tapered. The rest of the total, longer, path is represented as a line.

I've briefly tried to see if the calculated trajectory matches a dop sim. It does but you have to calibrate for scale once. With moving targets and shooters it looks all terribly chaotic so I left it out of this video.

There's a big blue shadow in the vision part crossing the scene. Not sure what that is. Either I forgot to switch off a light or its a very nearby ball. Or a ghost.

The targets where done in a dopnet with a force that tries to keep them together, so they never escape an imaginary force field around it.

I also had a bit of fun rendering a pencil look on the gray one.

https://vimeo.com/65116981

post-4013-0-14132800-1367330475_thumb.jp

post-4013-0-47458400-1367330567_thumb.jp

Edited by Macha
  • Like 1
Link to comment
Share on other sites

Thanks Edward.

Sensor Vops??? I'll have to have a look at that. 12.5 I suppose.

Edit: Nice nodes! I should have checked out the new features a little more carefully. I had thought about it this stuff a while before 12.5 and hadnt had acces to it until very recently. It's a neat point release with many nice improvements.

Edited by Macha
Link to comment
Share on other sites

For me, the deforming example doesn't work (errors) and the rbd one does nothing. Trying the nodes myself also throws errors. I'm using a Mac and I guess that's the reason.

If they use opengl then I assume sensor vops cannot be used on farms. The other problem I see is that we cannot directly use an image but have to save it out. Perhaps we could send it to cops with an expression but if I recall correctly that's not farmable either. So, you'd have to precompute the sensor results if you wanted to do anything clever with them (ie pipe them through image analysis) but then that would make it hard with dops work, unless you save previous time steps and work on them. Not sure how that would work in practice.

Edited by Macha
Link to comment
Share on other sites

  • 4 months later...
  • 2 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...