Matrix Games Forums

Hi,

It's an idea I've had this afternoon that I think could spark more interest in the CC series, and improve the in-game AI...

I was reading an article about AI in StarCraft (click here). StarCraft is a hugely popular SF RTS game. And what they (i.e. the developer of StarCraft) did recently was to open it up so that the game can be driven by an external AI engine. This is interesting because, as the article explains, quite a lot of research is nowadays happening in AI in games, and in RTS games in particular. So after having "opened" the game this way there was quite a bit of interest from universities from around the world. As a result, a competition was announced in which various AI engines would face off in the game.

So I thought that a similar thing would be interesting for CC as well... CC is a stable game engine, refined over the years - something valuable in a research environment. And the outcome of such a move could be (a) increased interest in the game, and (b) ideas for how to improve the in-game AI.

There )))

Best,
blazej

That's an interesting idea! I had heard something or other about the AI in StarCraft. I haven't played the new one, but I lost many hours to the first one.

Along those lines, I play another military game, a flight sim called Allied Force Falcon 4.0 (its based on the flight sim Falcon 4) The way the AI and campaign AI is handle is by dedicating it to another processor. Most computers now days are duel core and what not, and one processor controls the flight sim, the other one controls the AI everything from enemy planes to the ground forces. Its makes for what is called a "dynamic" (never plays exactly the same way) campaign.

I honestly think, if anything is to truly benefit AI in general, it is finding the best way to utilize this to its max potential.

But I agree opening the AI code to the community, would allow us the opportunity to tweak and modify it as we see fit. Much like how we mod all the different versions of CC. Not to mention developers do have a time table they have to meet, and they can only dedicate so much time before the game has to be released. Opening it up to the community would allow us all (the community is much, much larger than the dev team) to work it and learn it at our own pace. Much like we have finely tuned mods like GJS/Stalin/Oka/Bloody O,ect for CCV, we could mold and tune our AI in the same way. Making each AI "mod" to suit a certain play style, eventually combining them into a super AI (SkyNet?).

The main draw back I see with all modern AIs is that they are strictly reactionary. Meaning, it depends on the player to determine its course of action. If their were some way to code the AI to anticipate, or dare a say "imagine" (as best that something mechanical could) multiple courses of action and prepare for them, much like a human player would, then we would break down the first major barrier hindering AI.

Course on more technical terms (if you can call it that), I honestly don't think we will ever have a true AI, until we find a way to blend something organic with software. In my little dream world that is the next true evolutionary step for it. We can only make something so small, or use so much processing power, or utilize so much RAM and memory. If there were a way as I said to blend something organic with it, an artificial brain lets say, to give the AI the ability to leap beyond logic (which is our greatest/worst asset) to allow it to "think" for itself, rather than solely react. Of course then some idiot would want to hook it into our nukes and we all know how that ends.

Aw I have headache.

Hey SlyGuy, do I know you from someplace?

Well that's why I mentioned universities, those guys do some cutting edge research, just read the article I linked. They could most likely build an AI that would think ahead and the such.

Good idea with the processors, though.

Another thing: it's not about opening up the code of the AI, but rather exposing the game engine as an API so that peiople could program and plug-in their own AI engines. Then hopefully some of that innovation could find its way back into the built-in AI

Best,
Michal

When I played the original CC4 grand campaign to death I was only wishing for one improvement:

That the AI remembered my deployment scheme between battles! Often, after I "optimized" my setup for a particular map (MGs, AT guns) as the US, the Germans just got slaughtered over and over again in the same way) .

Had they remembered my deployment positions, they could have defeated my defenses with recon-by-fire quite easily, making for a much more interesting fight!

Best regards,
Thomm

ORIGINAL: Thomm
When I played the original CC4 grand campaign to death I was only wishing for one improvement:

That the AI remembered my deployment scheme between battles! Often, after I "optimized" my setup for a particular map (MGs, AT guns) as the US, the Germans just got slaughtered over and over again in the same way) .

Had they remembered my deployment positions, they could have defeated my defenses with recon-by-fire quite easily, making for a much more interesting fight!

Allowing the game to "remember" those kinds of things was the idea behind the never implemented map and player 'history' files (.hst) in CC3.



btw - you wouldnt still have copies of those excellent quasi-3D overview maps you did for CC4&5?

ORIGINAL: xe5

btw - you wouldnt still have copies of those excellent quasi-3D overview maps you did for CC4&5?

Short answer: No.

Although I am unable to figure out why. I do have the source code of the program that produced the map. Perhaps I concluded that I could reproduce the maps any time I wanted, and this is why I deleted the bitmaps themselves. But since then I changed computer environment and it may take some time to re-compile and re-run.

Best regards,
Thomm