standardhero

Cellular Forms

Thu 23rd Jan 2014, by Paul Hellard | Peoplestudios


 

 

Andy Lomas shows off his new friends, grown in his computer.

 

Andy Lomas has an understandable love of those old Victorian zoologist’s botanical drawings. The details in the parchment drawings are so full of wonder, down to the finest scale. Meanwhile, for many years at home and in in the background of his job at The Foundry, Lomas has created what he calls his Cellular Forms.

 

Emergent

“The studies of organic structures is enthralling, but when you can create digital cells from scratch from software in the virtual world on your screen, the visual possibilities are limitless,” he says. Lomas readily admits he doesn’t understand why these shapes end up the way they do. “This really is emergent stuff. I was hoping interesting things would happen, but it is purely unpredictable and is not what was expected,” he quips. “It’s great to have this at home because I can create in freestyle without limits. If I was creating stuff like this at work for a movie, it wouldn’t be what the director wants.”

 

In fact, when he is working at The Foundry in London, Andy Lomas is working the Katana team, together with Sony Imageworks, on lots of engineering, including a complete overhaul of the core engine that Katana uses. “This should make a lot of stuff a lot faster, as well as allowing users to write even more powerful plugins that we're calling 'Ops'”, he adds.

 

 

But, when he is at home on a weekend, he creates some intensely natural looking images, using a piece of software he’s also grown at home. “I have made two main programs for the generation of these forms. The first one is called ‘ParticleLinks’, which I use to simulate my digital cells and have as a render engine. And ‘SpeciesExplorer’ is the software where you have all the parameters and increments. This lets you do things like rate and categorize the samples run so far, then it uses that information to look for the transitional areas between things that are categorized together. You can cross-breed samples, or just pick an interesting sample and create new mutation variations around it. There are so many variables for gravity and radial forces and many others. The challenge is to find out exactly what this machine can do in the multi-dimensional parameter space.”

 

 

“I set the system up and let it go, and that’s the beauty of this. In a production, you have to create something that is controllable. You get the notes from the review session from the VFX Supervisor and the Director, saying ‘change this, fix that’, and you have to build something that is pretty much direct-able.” This is Andy’s reaction to that, in his free time, where he, as he puts it, “lights the wick and lets it go.” The creations and reactions to the surprising visuals have been amazing.

 

Cellular Forms uses a simplified biological model of morphogenesis, with three-dimensional structures created out of interconnected particles to represent cells. A number of internal forces affect the structures, including linear and torsion spring forces between connected cells. Additional forces repel cells that are in close proximity but are not directly connected. This imposes important physical constraints on the system, which gives global structural coherence to the resultant forms.

 

 

Cell division can be triggered by a variety of means, including accumulated nutrient levels. When the nutrient level in a cell exceeds a given threshold, the cell divides. Various parameters control effects such as how the plane of cleavage is chosen, and how both the parent and daughter cells re-connect to their immediate neighbors. Rules can also be adjusted for how nutrient is created, such as by being randomly created by each cell, by use of ‘reaction diffusion equations’ to stimulate different growth rates in different areas, or by incident light rays creating nutrient in cells hit by photons. Nutrient can also be allowed to flow to adjacent cells.

 

The simulation process is repeated over thousands of iterations and millions of particles, with typical final structures having over fifty million cells. Ray-tracing is used to render the particle data from the simulation into two dimensional images. The resulting artistic artifacts are high resolution images of final forms and animations that show the progress of the simulation over time.

 

 

Rules

Andy Lomas studied maths in an area that interested him, called ‘dynamical systems’, which is a discipline exploring how things can change over time. “This is where you may have a rule that you continually reapply to the problem,” he explains. “‘Growth’ is an expression that feeds into this dynamical system perfectly. This is the area of math that brings in the Chaos Theory.”

 

The Transition State is the point where the visual suddenly changes from one form to another. Lomas has been experimenting with the area where that tipping point is, with 20 different parameters. “One moment it will look like a cauliflower and then next, it has morphed into a worm,” he says. Lomas runs all this on a gaming spec PC with a GTX Titan card which has been running pretty constant since he bought it. He first ran the code on the CPU and it wasn’t fast enough, and a friend from NVIDIA persuaded him to go the GPU route, and now the ‘ParticleLinks’ has been re-implemented in CUDA.

 

“’ParticleLinks’ is what needs the real turbo charging as it's doing all the simulation and rendering, with over 50,000,000 cells in most of these sims,” explains Lomas. “SpeciesExplorer just has to pick parameter values and submit scripts for ParticleLinks to run, and I implemented it in Python and PySide.”

 

 

Variations

Lomas sees this as a way of testing something on the computer, ‘in silico’, as opposed to ‘in vitro’ in the petri dish.  He starts with literally four CG cells, which grow at a set rate, then they split and continue to grow. There are forces and interactions working on the sides of each cell as they grow and split. There are no texture maps, displacement maps or 3D noise functions. “It’s as simple as it can get, running to millions of cells, running on a GPU on a desktop computer,” explains Lomas. There’s a point after the creation goes past a few million transitions that it begins to take on a purely biological appearance.

 

“There are a lot of parameters set at the beginning, like the tension between cells, the torsion spring, which decides the amount of bending it goes through, as opposed to the coiling action,” Lomas says. There are rules governing things like when a cell splits, for which direction it splits in. This is governed by how much surface tension there is at a particular point. These settings don’t change at all during the simulation.” Lomas had some code called a ‘reaction diffusion equation’ within the SpeciesExplorer that would make natural areas appear that would show lower growth rates in a random way, but then even without that diffusion system, the cells showed that randomness, ‘naturally’.

 

 

Colors

When it comes to colors, Andy isn’t all that fazed about putting particular colors into the images, for their own sake. However, influencing the growth patterns of the cells, dependent on the color of the light particles visible to each cell may be something he would experiment on in the future.

 

On the Cellular Forms page on Andy Lomas’ website, there are images of cells that have been grown with an attraction towards light where the luminence has only been coming from the top of the picture. In these CG growth experiments, these cells seem to adopt the habits of natural forms in more direct ways. When the light comes from the top, his creations naturally end up taking on the form of plants. And this isn’t because they are training to grow towards the light, only that they grow when they are exposed to the light.

 

“One thing we could do is visualize the stresses and forces on the edges of the cells could be colored. Showing the stresses and movement of nutrients around the surface, almost like gravity. There are all sorts of things you could include in this experimental tool kit,” he adds.

 

 

He first showed this new cellular work at a science symposium in Brighton in October 2013 and there has been great interest in his original Aggregate video creations from 2006.  Despite the request for medical visualization projects, Andy admits he wouldn’t really be able to use this system to recreate particular molecules of cells, and the methods with animators using Maya would probably be less trouble. But who is to say what the future brings? One might imagine a whole strain of modern day Tamogotchi, or a new wave of virtual Sea Monkeys, able to live in your system, endlessly growing and morphing into something purely beautiful and chaotic.

 

Andy Lomas’ new Cellular work is showing at a solo exhibition, Morphogenetic Creations at the Los Angeles Center of Digital Art (LACDA), opening January 9 to March 1, 2014.

 

 

Related links:

Cellular forms

Aggregate

LACDA  

Discuss this article on CGTalk

Writer: Paul Hellard 


blog comments powered by Disqus