Getting this far actually took a few sittings. I had started this sketch, but had gotten stuck with a simulation that was getting extinguished quickly.
When mentioning this to the friend who told me about the paper in the first place, he also sent me a link to the author's original BASIC code -- So I made a separate sketch (not included) that ports that to JavaScript to learn how it worked. I learned a few things:
- The shell isn't rendered every iteration, but rather every few iterations
- A couple parameters from the paper weren't actually used for the Olivia Porphyria shell. For example, the inhibitor production rate should be 0 in this case (which was one main cause of my incorrect implementation)
- delta time and delta x were both assumed to be 1. I find this unusual, I'm used to deltas being a small fraction. But it keeps it nice and simple.
- The original paper did some clever indexing to avoid allocating a second buffer.
After that exercise, I went back and updated my implementation. I used more descriptive variable names since there are many parameters. I also used a double-buffering approach since memory is cheap nowadays.
After that, I started moving the parameters in to a SeashellParemeters
struct with defaults that indicate "no change in chemicals". This way, I
can experiment more easily, and save parameter presets as constants.
Possible Next Steps:
- Experiment with the parameters
- Experiment with the initial conditions
- Experiment with varying the parameters over time
- Try the other activator/substrate model for comparison
This morning I spent a little time experimenting with the parameters. Still don't have much intuition of how the patterns will look, but made a few interesting designs:
LUMPY_STRIPES:
BLOCKY:
One thing I've noticed is that the boundaries tend to be rather pixely. The
original BASIC program did a simple thresholding of the activator value when
picking colors. it might be good to introduce a smoothstep function and use
that to blend between the substrate and activator.
Another thing I'm wondering is what would happen if I mixed in the color when the hormone value is large? Might be helpful to understand the dynamics here.
Possible Next Steps:
- More experiments with parameters
- Try the smoothstep idea
- Try visualizing the hormone levels
- Adjust the image dimensions for a shell texture (which would be longer) and/or standalone images (I like making trading-card sized artworks)