EAL: Evolving Artificial Life

This is an old ALife program I wrote years ago just for fun.

Each coloured square on the grid is a single creature. Each creature can examine the colour of grid cells directly in contact with it (up, down, left, right, and the diagonals).

The program starts off with a bunch of randomly-generated creatures. Most of those die out very quickly, but a few manage to reproduce.

Some parts of this video are showing the program running at regular speed. Others show a 'time-lapse' mode that runs through simulation time much faster. There's no visible marker to indicate which mode is active, but you can usually tell by the amount of change from moment to moment. This video is not a single run, but a combination of several independent runs.

The creatures' genetic code contains a list of rules. Translated to English, the rules would read something like "if there's a green square above you and a food square on the right, move to the right". There are more conditions, more options, and more actions available than just that example - but that's the general idea.

There's a kind of 'gravity' that pulls everything down in the grid. Creatures can climb up and around each other to overcome gravity, but if unsupported, they will fall.

Food pellets (dark grey squares) fall into the grid at the top.

Creatures can reproduce either sexually or asexually. Their rules include actions like moving, eating, reproducing asexually, trying to reproduce sexually with a neighbour, sharing food with a neighbour, and things like that. Some other genes help determine who is willing to mate with whom, who will win in fighting encounters, and so on. The winner in a fight eats the loser.

Since food falls from above, 'species' that climb up high or build colony structures that choke off the top of the grid are rewarded well (just by getting the food, I mean - I there's no programmed reward for any particular behavior). Those stuck below them often have to get food by attacking others.

Predators can appear and wipe out all of their prey, and then go extinct themselves. All kinds of interactions like that can occur in the grid.

Creatures of the same type (or closely related) have the same colour (this doesn't mean two creatures of the same colour are necessarily closely related - but usually it's true).

Creatures can die off just from lack of food. When creatures die like this, they are 'recycled' as food pellets that fall from above. The total 'energy' budget (or food-pellet-budget) remains constant.

Category:

Film & Animation

Tags:

• artificial
• life
• 2D
• evolutionary
• computation
• evolution

[ehsanul]

this was a captivating watch.. i should really try to make an a-life program of my own.

[kjlg74]

Thanks!

Yeah, absolutely, give it a try. They're fun to program and fun to watch. If you have any luck with it, and if you post a video animation, please let me know. I'm always interested to check out new ALife and evolutionary computing stuff on youtube :)

[pbazant]

Great program! I am interested in getting a copy of it, if possible. Is the brain of the organisms Turing-complete? I have written an evolutionary simulator called Evoversum: see watch?v=nPkZvyVDWJU 

[vamori]

Is there a download for this? it sounds awesome.

[kjlg74]

(sorry if this is a duplicate comment - the previous one didn't show up).

I'll send you the code :)

[TheReasonWhyGuy]

LOL

I have time to record something for you ;)

Also I could get a first hand look at it, but if you would rather do it yourself I understand. :D

[kjlg74]

:D Oh. Well, just keep shrinking the grid until it runs at a good speed. This one is 50x50, I think. Come to think of it, that's minuscule :P I should try 1000x1000 sometime. It would run about 400 times slower, but if I ran it for 24hrs I'd get a 3.5 min clip... ah, but I have no time for that. At some point, though.

[TheReasonWhyGuy]

"yours should run faster as well."

I doubt it since the thing I used to code isn't exactly wired for speed :/

But thanks for the encouragement. :D

[kjlg74]

Good luck! There's nothing overly complex or fancy in EAL, so the chances of duplicating it (not the the letter, of course, but in essence) are good. Plus, without the constant flood-fill-style algorithm I used to do the gravity stuff, yours should run faster as well.

[TheReasonWhyGuy]

Hmm, interesting.

You may have just inspire me to make a version of this myself. Although I'm not sure If I will do as well as you. I will definitely try some new commands and maybe make the creatures more versatile.

Also with the idea I have I might make it so you can look into the creature genome and see what its all about.

I at least hope to copy this. Although I think I'll do mine top view, no climbing just chasing. But more pixely. :D