ConwayLife.com

I've admired CGOL for a long time but am very new to actually programming Cellular Automata. That the simple set of rules and bare minimum of cellular attribute(s) (there is of course precisely one attribute, existence and its negation) give such a startling variety of patterns and interactions, is an impressive testament to JHC and his skill and zen / tao - like sense of balance. ( I say that in complete ignorance of Eastern Religions other than through crass Western popularizations)

However, having a background in physics, rather than pure mathematics, after just admiring it for a while, I got thinking of how CA could: do real work in physical system modeling, computational modeling, etc, if they had other attributes. I don't think this is the time to list the ones I'm thinking of, but clearly they would be attributes that the rules would be using to calculate the subsequent state. I want to correspond with anyone who has done any thinking or coding in this area, or if anyone has any references. I realize that there are a lot of resources out there, but most of them seem to be academic papers, and few of them include source code. One big exception to the dearth of research in this is CAPOW by Rudy Rucker and his associates and students:

http://www.cs.sjsu.edu/~rucker/capow/

http://www.rudyrucker.com/capow/capowhelplong.htm

Please follow up to this post if you have any interest in these topics, or e-mail me at:
chad AT alumni DOT unh DOT edu. Thanks a lot.

Wireworld might be a good place to start. It has pretty much all small logic gates, and interesting/useful patterns can be constructed in the rule, such as a prime number calculator, which actually displays the prime numbers on an LCD-esque display.

If you're interested in modeling physics and computation with CA, keep in mind that there is a tradeoff between states and area. For example, CGoL has 2 states, while Wireworld has 4. Wireworld requires double the memory for any area that CGoL does, but CGoL almost always requires more area for logic gates than Wireworld. Wireworld signals move at lightspeed, while CGoL signals move at c/2. (at least the practical ones. See the lightspeed telegraph included in Golly's signal-circuitry folder) However, I think that Wireworld might require a bit more processor power than CGoL because it has more states. (Can someone confirm this?)
Another thing that is important is constuction ability. In Wireworld, blank cells always stay blank, ruling out any possibility of contructors. CGoL, as you probably know, is capable of construction. Codd is similar to Wireworld, and it's capable of construction. However, it has (I think) 14 states and takes up more space than Wireworld. It requires 8 times the memory of CGoL, and constructs things a good deal slower, but it has simple lightspeed signals and is generally more compact than CGoL.
Creating a new rule to model physics would be interesting. How would you handle collisions? Particle speeds?

Thanks folks. I'm very patient and getting better at incremental development. One of my colleagues who is not a CGoL guy at all has had a lot of success at creating running models of systems with dx/dt & d^2 x/dt^2, gravitation, and handles collisions in a few, principally elastic ways. I have plenty more work to do before modeling E / B fields... It would be great to have adsorption to facilitate aggregation. The Lattice Boltzmann work is fascinating to me. I'll look in again soon. No prototypes yet...

theDragn wrote:However, I think that Wireworld might require a bit more processor power than CGoL because it has more states. (Can someone confirm this?)

Very slightly more, yes -- at least I think so. In practice, Golly's Hashlife algorithm tends to minimize the processor-power difference, at least when patterns have some regularity to them.

People almost always seem to run fairly organized patterns in WireWorld... so once the rule is loaded and HashLife has had a chance to get up to speed, every subpattern has already been seen before. Once everything is getting looked up quickly in a hash table, I'm not sure you'd see any difference in calculation speed at all.

Now, Wireworld has only four states and relatively simple state transition rules. I suspect that a random chaotic pattern in a 255-state rule would run a lot slower than a similar random pattern in a two-state rule, because with 255 states the HashLife algorithm would be swamped -- no hope of ever seeing any large subpattern twice in a row.