AI, Where it's at! - Page 3

Dominia · 06-12-2007, 10:09 AM

If you just recently took that AI class (in the past 3 years), then yeup that's it. Early version sounds like. =)

Not much to look at, I know lol

--Jason Hooten

ronnoc10 · 06-12-2007, 07:42 PM

Heh, that sounds like an amazing amount of fun! When'd you say you where going to grace us with this peice of art?

Quote:

'm not sure but I think we may have used this a a example in one of my AI classes. Or at least a very early version of it.

You created robots who wandered, Put a target location and let them aim for it.

You would smite the ones who did not do the right thing and as a result only the smart survived in the 'gene pool'

You could build walls and stuff too to make it more harder.

Evolutionary programing? or am I once again missing the point?

Dominia · 06-13-2007, 12:17 AM

hmm since you didn't play the game or follow the links to wikipedia I put up I'll describe in further detail how the learning works so it will hopefully be clearer for you =)

In training mode on your hud is a toggle bar which sets how much or little you can reward (or punish) a particular action or result. One of many sliders is "Aviod Fire". If you if you set the slider to high reward the nero agents (a) get rewarded when they see gun fire and are not being hit, (b) get punished when they do get hit. If you wanted 'meat sheild' type robots you of course would set the slider in the opposite direction to reward getting hit.

As they get rewarded and punished, thier NN's get fine tuned accordingly. So just like a pet, they learn by analyzing situations and applying previous data to new problems.

Now what Hagnasty meant by smiting agents, is if agent has randomly wonder off by himself and is not being productive in learning the exercise you have prepared, you can smite him. This will not result in any negative or positive feedback, it will only respawn him at the spawn point.

There is a chance that Hag is not talking about nero at all as well, I just assumed so because of the smiting, gene pool, and wall building references . I can imagine many games have used those same terms though (i can think of 3 off-hand lol) so who knows

Another really neat AI research being done by a company called Novamente. They've got a couple of cool SeedAI type engines coming out in the next couple of years that can be interfaced with games, UI, robots or w/e. Very neat stuff. The next couple of years for gaming will be interesting.

--Jason

HagNasty · 06-14-2007, 09:09 AM

Yea I think I examined a different one. I searched the one you speak of and even version 1.0 is more advanced. The one I played was top down and when you smite them it removes them from the gene pool completely. There where no trees just flags, walls and robots.

ronnoc10 · 06-15-2007, 06:26 AM

Oh. I didn't see those links the first time around...

Dominia · 06-17-2007, 07:50 PM

Hesham, that is very cute, looks like something a friend of mine is working on at the university (Robot Soccer, though it looks like it's a different tournament then yours (yours being a little more professional/industry based, idk).

About the open source version, it won't be ready to release at least for another couple of months, sorry =(, and even then it would only be a research tool, not the industry interface. If you have any questions though about feel free to email me and I'll help as best I can.

AI aside, that Eco-Be is an amazing little piece of hardware

, that has to be fun to program for. On the article, good summary of everything AI in gaming atm though referencing MMO's to relate human and AI interaction seemed out of place to me at least. What really got me was the line, "World of Warcraft, one of the oldest [mmorpg's]," lol not by along shot but I it does get the point across.

It is good to see info on gaming coming up in new places on the net, the more we can market gaming as a profitable and viable venture the better off we all are =).

--Jason Hooten

ronnoc10 · 08-02-2007, 08:26 AM

http://www.gamasutra.com/view/featur...ithms_for_.php
Hey look, desireiblity based AI.

Or as he calls it 'priorities'

Gshonk · 08-02-2007, 06:37 PM

Figure i should jump back in here since I am doing the AI for the game I am currently working on. I made a 2d top down game 2 months ago and the AI was super simple in that, it was just based on the players position and updated by it. But i threw some cool stuff in there so it was like a swarming affect(aliens look better when they swarm). But the AI I have to do now is just a little bit more difficult since the game is a side scroller (dun dun duhhhhhh). Ok so it still should not be hard we basically have a few kinds of AI planned. A basic AI that moves the guys back and forth depending on if they are near an edge or not. A finite state based AI with like Patrol. Attack and Flee states. And the boss AI with just a few different attack states. I start working on the code tomorrow so I'll let everyone know how it goes. AI FOREVER.

HagNasty · 08-08-2007, 11:34 AM

Quote:

Originally Posted by ronnoc10

http://www.gamasutra.com/view/featur...ithms_for_.php
Hey look, desireiblity based AI.

Or as he calls it 'priorities'

Yea basically this is reinforcement learning ^.^ long term goals with weights choose the best one to achieve the goals.

Gshonk · 08-08-2007, 01:44 PM

http://www.ai-junkie.com/ai-junkie.html i love this website.