View Full Version : AI-Junkie-Competition 2003: Emergent Behavior

07-07-2003, 12:22 PM

08-07-2003, 06:42 AM
If you're interested, I made a Delphi version of his Minesweeper AI demo. Doesn't work as exspected, but if you can find the bug, all the power to ya.

30-07-2003, 07:45 AM
I've been trying to think of what to do for this competition. I'd like something to do with animal packs.

- A wolf simulation showing the teritories the wolves use and the forming and breaking up of wolf packs. Including obstacles that change the shape of the territories and the availability of food with in the range that defines the size of the pack. Rules - Defend food, Pack size Limited to food, chase other wolves if in range
- Pretty much the same but for Dolphins
- Tribal mechanics for African tribes
- The nesting habvits of birds - specifically the gannets on the small antarctic islands - but I've found it difficult to define rules.

The the other thought I had was one based on the movements of the planets. ie Velocity and Gravity. This will allow you to enter various random bodies with a gravity value and a velocity and direction, then by applying the rules of velocity and gravity we should end up with a solar system :)

anyone got other good ideas

Anyone have any other ideas?

30-07-2003, 09:48 AM
My cat's breath smells like cat food.
AI is a bit over my head atm. Maybe next week... :drunk:

30-07-2003, 11:30 AM
I was thinking of the behaviour of a night Butterfly which moves from a bright surface to another ... But I haven't yet the skills in order to make such a sim :) (So it'll be for the next competition :P)

30-07-2003, 12:25 PM
You would need to work out the rules for the moth to move between lights.

Typically the rule set wouldn;t have random values in it so you'd need to make clearer rules.

Possibly never allow the moth to look behind itself so that it will not just fly between two lights all the time.

1. Moth flies to nearest light
2. When moth reaches light - choose the closest light other than the one it just came from. - Go to 1.

This would in fact be quite interesting to watch and see the clear pattern that emerges as lights and moths are added.

Any objection to me coding this for you?

30-07-2003, 12:46 PM
Lol, no objections :) As I said, I haven't got the skills yet :)

But could you send me the sources when it's done ? I'd like to take a look because AI is a topic that interests me a lot ^^

Thanks a lot :P


31-07-2003, 06:30 AM
My 7year old daughter had a choir event last night so I was forced to sit still for 3 hours.... in other words I had a lot of time to think about this :)

Anyway. Why do you think you do not have the skills to do this? AI is just another way of programming and does not need any specific skills.

I'll do the coding and send you a copy. It shouldn't be more than a couple of hundred lines of code anyway.

31-07-2003, 10:47 AM
By skills I mean that I would need some knowledge I haven't got for the moment. I looked on AI Junkie and on others web sites about AI but ... I ain't got the time(for the moment) to learn that knowledge. Ok, some AI systems aren't very complex, as the FSM systems and others, but I mean NN and Genetic Algorithms aren't that easy to understand ...

On the top of that ^^ You add that I'm french and that I didn't understood the whole definition of an Emergent Behaviour ... You know then why I say I haven't got the skills yet in order to program this ...

But, if you give me the code I'll be able to see how you did it, and how it works. So i'll be able to learn some knowledge lol ^^

I don't know if I'm clear enough :) But anyway thanks in advance for the code :P


31-07-2003, 06:00 PM
I'm not quite sure if this is really on topic, but still. I have made an AI Unit based on Fupster's Neural Net and Genetic Algorithim model. I've used it in a few programs and gotten it to work with digital functioning however I have not successfully gotten it to act as wanted in analog mode.

If either cairnswm or Avatar (or anyone else for that matter) would like to have a poke at it see what it's made of, I'd be glad to send the code. Source wise it's not very big actually. That and I could use some help in testing and debugging it.

Wish i could have been able to enter into this contest, just don't have the time to hit as many books a I'd need.

31-07-2003, 06:40 PM
I would take everything you could send me ^^

Just take a look in my profile :) There's my mail :P

Thanks a lot !

31-07-2003, 08:07 PM
You sure? It's not letting me see it... you could drop me a line instead. raproductions@hotmail.com

01-08-2003, 06:54 AM
I must admit that I really do not understand NN or Genetic Algorithms very well. Maybe one day :)

Anyway back on topic this competiton is about Emergent Behaviour. Based on the description on the competition web site it effectivly defines it as a rule based system. Rule based systems basically work on IF statements and all or us can use them.

I'll start work on the Moth simulation soon. I'll post progress as it happens.

01-08-2003, 08:52 AM
You sure? It's not letting me see it... you could drop me a line instead. raproductions@hotmail.com

If you click the 'email' button under a post you'll be able to email a member via the site (no attachments though :( ). All email addresses are hidden now after requests from several members to keep them private.


01-08-2003, 08:32 PM
cairnswm: NNs & GAs aren't that hard of a concept to grasp given a little bit of attention to understand it. Actaully Neural Nets are quite easy to grasp if you know a little about digital electronics. All NNs are is one big interconnected circuit. GAs in theory seem simple enough, and I *think* I understand the concept fully, but then there is that grey area on how it used... So far, I've found Fupsters tutorials to be the most useful reading materials for learning about this. However the fact that he uses C/C++ to explain it in code loses me in the translation. It would be great if someone could rewrite or translate his tutorial's code to Object Pascal, I think it may become clearer on how to do this in Delphi.

BlueCat: Why not make the hiding of emails optional, but have the default set to hide it? That would seem to make more sense to me.

02-08-2003, 01:02 PM
I'll add them to my 'one day I will...' list :)

10-08-2003, 07:12 PM
I've completed AIMoths:

Screen Shots:

Source Code Only

23-09-2003, 12:06 AM
I'm going to be doing a super, mega, neato flocking-like behavior example. heh.

-3d ogl a must for the eye-candy
-I think it will be the most advanced display of this type of artificial emergence (both visually, with the graphics, and with the concepts).
-I hope there are more entries from this group (i'll provide some basic source showing very simple flocking)
-As Alimonster mentioned in the Artificial Life Forum, http://www.riversoftavg.com/flocking.htm has a working example (albeit muddied with the components and at its core only a very simple implementation)

I hope we have some contestants - ask here on dgdev or email me if you have any questions (tclinecompappii AT hotSPAMmail DOT com)


23-09-2003, 05:03 AM
I will be doing an entry. I first need to get a game done for the other Competition (Deadline 21st October).

I am not sure what sort of rule based emergent behaviour to make...