Archived

This topic is now archived and is closed to further replies.

Corvus

...or not alife?

Recommended Posts

Corvus    122
Actually I am looking around for information to be able to create an environment where creatures run around, eat and reproduce themselves. Graphics are not important at this moment. I guess pathfinding is important and some simple sort of genetic algorithm... but I don''t really know where to start, or which ai topics i will need for this projects. Any help? +++ Corvus +++

Share this post


Link to post
Share on other sites
Corvus    122
Yes I know the game, I''ve been playing Creatures 1 for a while. I''m really impressed about the way they reproduce (genetics I mean).
So in a first phase I want to write algorithms to try to approach a bit of the logic behind that game (no graphics).

So my question is: where to start? I am a professional programmer, so technically there isn''t a problem .

+++ Corvus +++

Share this post


Link to post
Share on other sites
RPGeezus    216
I agree with Fup.. His website has some good articles.

If you go that route, what you should invesitage some form of Evolutionary Programming.

You can use either a state machine and hope for the best, or write you''re own pseudo-language (which is probably more effective). There is a lot of literature available on this subject.

In ''96 there was a big 3D online project that had all kinds of artifical creatures in it. A big predator - prey thing. Too bad I can''t remember the name. Anyway, it kind of went loopy because the predators would end up cornering the prey and then picking them off until there were none left. To avoid a situation like this, I would make sure that your creatures:
a) Poop.
b) Don''t like being around poop (unless it''s prey poop).

This way, it should prevent them from staying in one place.


Good luck,
Will

Share this post


Link to post
Share on other sites
Timkin    864
The creatures architecture is exceptionally complex and combines aspects of evolutionary programming, artificial neural networks, neurochemistry and physiology. It''s not something you should tackle as a first attempt at an A-Life simulation... heck, most A-Life researchers wouldn''t dare tackle it either!

fup''s suggestion is quite reasonable, although that too may be more than what you need for your simulation. Start out simple... define the sensory inputs you want your agents to handle and the actions they can perform. Then choose a method for mapping inputs to actions: either a state machine, classifier system, ANN, etc. From there you can decide if you want individual agents to learn during their own life time (reinforcement learning) or the species to learn through evolution (evolutionary learning).

If you want to go down the evolution path, then you need to choose between simulating mating between agents (perhaps when they are close to each other and ''in heat'') or whether you take the hand of god and run a genetic algorithm at regular intervals.

There''s plenty of literature online about this sort of thing, so you shouldn''t have any problem with finding something to read!

Good luck,

Timkin

Share this post


Link to post
Share on other sites
Corvus    122
Timkin: don''t worry, I wasn''t planning on making a new creatures game .

I have decided that the first step in my project will be movement and pathfinding. I think this will keep me busy for the next few weeks (depending on the time I am going to spend on it). After my agents can move around a bit, I want to make them more intelligent eg. let them look around for food and friends.

Suggestions are always welcome!
Thanx for the response so far...

+++ Corvus +++

Share this post


Link to post
Share on other sites