[Modpack] Self Organizing Systems 0.3.1

[Dokimi]

The evolving vine (the purple flowers).

The number shown controls the plant's life history strategy. It must trade off a long life versus fast growth (that is, it can use an r selected strategy, or a K selected strategy). Each time it reproduces this number randomly goes up or down a little bit. Natural selection does the rest.

[azekill_DIABLO]

infinite GOL in minetest.

[Dokimi]

Massive update: 0.3!


Added a bunch of things to the current mods I hadn't got round to earlier:

- A cellular automata that plays music... I mean "music" (has a few sound files you can play around with, if you'd prefer hypno-drone to demented clown piano!)

- More self replicating tools: a tunnel, "sinkhole", dome, platform, and ladder. You can now grow entire cities with these things.

- Selfrep_doomsday has been brought into game play: now it is an intense survival game. The world has risky labs that can accidentally release doomsday devices. But... you have "protector" blocks with which to wage your war against the machines. Also has two new doomsday devices: "Flash" and "Chaos".



And.... We have three new mods:

- "Autosky": a weather simulator. (unfortunately a little "heavy". Has a habit of doing either nothing, or super-storms that grind my computer to a halt.)

- "Neuron": a brain cell simulator. It's like Mesecons... but with epilepsy.

- "Evolve": the real star of this update... an evolving autonomous mining bot. By far the most complicated thing I have programmed so far.

...

[Dokimi]

Autosky:
Most of the action is invisible. But I've included colored nodes too, so you can see what is going on.

The transparent blocks are moist air. Either warm (red), or cold (blue). Evaporation, temperature, altitude... all control when air changes between states. Warm air rises, cold sinks... In the right conditions they form clouds, rain etc...

Add it all up - weather.

[Dokimi]

Evolve:
By default this bot mines coal, and leaves behind coal blocks... but you can set it to do all sorts of things (build mountains, convert the oceans to gold...). What ever challenge you set, it will use evolution to find the best strategy for success.

It gets a little involved...

The bot classes the blocks Above it, below, and to one side. It uses this to generate 64 unique situations. Each situation then has a gene. Each gene has several alleles. The alleles decide what action to take in that situation. e.g. AAA = MB means when in midair, move below.

Bots then get a score as they go along ( # target dug - number of moves). Every few minutes winning bots are selected, mated, and new superior genomes distributed to all bots.

And it works! Bots really do evolve strategies, even coming up with ideas I wouldn't have thought of myself.

The menu for a bot:


The most successful genome, used for mating (can be looked at from a computer block):

[azekill_DIABLO]

It lags no? this air system? it just sounds awesome to be honest, but i don't see my PC running that.

[Dokimi]

Small update, 0.3.1

Added a 2nd bot to Evolve - now you can experiment with co-evolution.

I have been trying to create a ecobots style mobs api (the current block animals are a bit limiting). It is including here (only adds a fish, in creative mode). It is unfinished. I have decided not to pursue it further - I add it only because someone else might want to take it further.

Likewise, I have been tempted to try and create an Ecobots api. It deserves to be taken further (I keep wanting to add species to it!). But I have decided I am not the person to do it. I am only good for experiments and conceptual things. Not coding, or admin.

I only have one other mod I may try for the SOS modpack - exploring multi-cellularity. After that I think I will close the book on SOS.

[Yvanhoe]

This looks like an awesome collection. I added that to the bookmark of mods to toy with for the educational game I am working on.

I would have liked to use cellular automaton to teach biology and plant growth. The idea would have been to either scale down the world or scale up the trees and see nodes grow as roots, stems, bark, leaves, etc... I'll read the docs but I suspect ecobots may be up to the task.

[Dokimi]

...I would have liked to use cellular automaton to teach biology and plant growth. ...

Some of these are broken now, and all are very badly coded (I learnt to code making these!). And some never worked well to begin with!

My recommendation for education:
1. Start with Cellular automaton. These teach self-organization at it's most simple level. Understandable, yet still fun. They are very easy to code. Maybe re-doing it with node-timers would improve performance.

2. Then move to something mid-level complicated e.g. like self-rep doomsday. Some of these include higher levels of complication e.g. positive/negative feedback loops. More exciting, but harder to understand.

3. Only then do something like Ecobots. This is extremely complicated. It would be hard for a student to understand how it works without trying something more simple first.

Check out https://github.com/DokimiCU/Ecobots2. It works and is better coded, but got very weird, so I never pursued it much further. This is close to being multi-cellular. Real multicellular growth is very hard to do (e.g. bark, roots, etc). Very complicated, and also all of these run into big problems with computer performance too. That is why I suggest starting simple.

[Yvanhoe]

Thanks! I'll check it out!

My (admittedly ambitious) idea is that this game would provide several paths to get at this point.

People interested in biology would arrive there and just learn the function of different parts of the plant, observing it grow, learning cell differentiation only relies on local conditions, that plants get nutrients through roots, use photosynthesis for energy, etc...

Another path would be through programming. I already have some mesecons exercises/mechanics. I am unsure about how to add luacontrollers or vbots to is, but an idea would be that after a while, they get introduced to cellular automaton and allowed to program some.

Then the (complex) model for plants would be presented to them.

Maybe (crazy plan) we can even introduce some notions of genetics, epigenetics and evolution.


Yes, the model would be pretty complex and the organisms big in terms of nodes' size, but I am thinking about providing a relatively small world to explain biology and ecology. Maybe something styled after the biosphere 2 "dome": An enclosed, somehow controlled and monitored ecosystem. That would allow things like running ABMs on all air nodes, allowing them to contain humidity, CO2 levels, spores... Something impractical in the typical MT world.

[Astrobe]

Rnd's trees ( viewtopic.php?f=9&t=13451 ) might be of interest to you.

[Yvanhoe]

Oh, Rnd's trees are close, but not complex enough for my goal. Thanks for the pointer, I'll look into it when I reach that point in the dev. I still think I'll need cell automatons but let's see.

EDIT: Actually the code is pretty straightforward. It is just two pages of code! Makes me think about how to do it elegantly...

[Astrobe]

Oh, Rnd's trees are close, but not complex enough for my goal. Thanks for the pointer, I'll look into it when I reach that point in the dev. I still think I'll need cell automatons but let's see.

EDIT: Actually the code is pretty straightforward. It is just two pages of code! Makes me think about how to do it elegantly...

For that sort of tree, your saplings could be like books carrying genetic code.

[Yvanhoe]

Exactly! Ideally I can find a syntax that allows complex behavior with very simple rules:
- if air -> generate protein A
- if water -> generate protein B
- if light -> generate protein D
- if protein A -> generate protein C
- if protein D -> turn into leaf and generate protein E
- if protein E -> create stem cell on the side

etc...