AI

[Arckon]

I have no idea on programming so following comment will probably show my ignorance.

Just listened to Dave O'Connor's presentation on Creative Artificial Intelligence.
Considering the lack of financial return for a lot of the wargame type developers and considering the quality of the AI in the Panther Game series my question is.

Would there be any financial benefit in bundling up an AI engine to sell to other developers (just like the better companies do with their graphics engines eg: the unreal tournament engine) or is the AI still quite specific to the Panther Games series. Seems a lot of the principles discussed in the interview would benefit, wego and igo/ugo products as well.

[Arjuna]

Arckon,
 
I agree that the principles and approach we use would be of benefit to other games. Yes it could be done but would require major surgery on someone else's game and would require a major effort on our part. Neither would be cheap and let's face it most of the AAA title games don't want to invest in AI.
 
Let me make it clear that you simply cannot just bolt on an AI engine. It doesn't work like that. It has to be fully integrated with the UI and mapping code. The best way to do it is to build it into the game from the ground up.
 
But hey, if someone wants to pay us a sizeable amount of money and it's for some project we're interested in, then yeh we'll consider it.

[Peter Fisla]

ORIGINAL: Arckon

I have no idea on programming so following comment will probably show my ignorance.

Just listened to Dave O'Connor's presentation on Creative Artificial Intelligence.
Considering the lack of financial return for a lot of the wargame type developers and considering the quality of the AI in the Panther Game series my question is.

Would there be any financial benefit in bundling up an AI engine to sell to other developers (just like the better companies do with their graphics engines eg: the unreal tournament engine) or is the AI still quite specific to the Panther Games series. Seems a lot of the principles discussed in the interview would benefit, wego and igo/ugo products as well.

I have been professional developer for 11 years (though now I work as business analyst in telecom industry) using C,C++, MFC and STL. I have also for the past 8 years been working on computer wargames has a hobby. First I started with making editors for Close Combat 1 then made my how hex engine to make Europa:Fire in the East type of operational wargame (3 years+) but without AI (how do you make an AI that works, with 3000 units in the map?) I'm currently working on computer version of ASL - based on ASL Starter Kit rules and as of this moment I do have AI that works pretty well (2 years+). So based on that here is my option: The short answer from my point of view to your question is no. And the very simple reason is that AI makes decisions based on its environment it fights in and the quality of the AI is in the details of the implementation. Making a good quality AI that just works with one game engine it is a huge amount of work (testing AI is very hard work as well), lots of time and lots of money. It has to be also optimized for speed and memory. There are simply too many variables to keep track of that you could use to make a generic AI smart enough that could be very well used for type of wargames. It's already hard enough as is to make an AI work just with one engine. A good example is Panther Games company. Dave and his buddies were working on the AI engine since 1995 so you can imagine how many years that have been working on it, tweaking it to make it what it is today. It would be extremely difficult to make a general AI engine that would be able to be as smart as in the airborne assault series and yet use it for say operational wargame series say SSG DBII engine and at the same time useful tactical wargames like say Close Combat series.

Peter

PS: Here is a screenshot of my ASL engine project that I'm currently working on.

[Peter Fisla]

Btw Dave, very nice AI presentation on your web site (MP3), currently listening to it; cool stuff!

Thanks Arckon for letting us know about Dave's AI presentation.

Peter

[JeF]

Peter,

In case you missed it, full presentation with slides and so on can be found on The Drop Zone. Nevertheless, not a lot of artist rendition of the object model of the engine. [;)]

Cheers,

JeF.

[Arckon]

I thought that would be the response otherwise would have been done already I suppose. Pity though, it would be great to see a company from this area of the industry find a way to actually make a decent profit doing what they enjoy.
 

[nicdain]

Peter, nice to know that you are working on a computer version of ASL. Is it the so called XASL that I have seen mentioned in other forums or something different?

Where can I have info on the progresses of this project? I am very interested!
Thanks

[Peter Fisla]

ORIGINAL: nicdain

Peter, nice to know that you are working on a computer version of ASL. Is it the so called XASL that I have seen mentioned in other forums or something different?

Where can I have info on the progresses of this project? I am very interested!
Thanks

Hi nicdain,

Yes correct the project is called XASL. Since the project is still heavily in development I don't really have site dedicated to it (don't want to be asked every day if it's done yet). The status is that I have the American/British/German/Russian nationalities in the game with all the squads/half-squads/leaders and all the support weapons and about 95% of full ASL ordnance for the mentioned nationalities already implemented. I have finished working on the first serious look at AI this year (AI in XASL is based on Solitaire ASL module); another serious look at AI will be given once I have vehicles implemented. I'm currently doing testing for few months with some of my friends (ASL players) and then I will be focusing on vehicles and artillery support. Further down the road I would like to implement dynamic/static campaign modules. Currently, scenarios can be fully played against AI (AI has different levels of difficulty). XASL AI is "generic", meaning AI is not scripted or tied to any particular scenario (scenario files only contain map data, order of the battle information and other scenario attributes like scenario date etc). I'm basically using the same design principles as Dave outlined in his presentation. XASL AI makes dynamic decisions on it's own depending on given tactical situation in a scenario. The AI behaves differently if it's on attack or if it's on defense. Scenario editor will be available upon release. I currently support about 35 ASL maps (note from ASLSK rules perspective + some features from full ASL but no support for hedges or walls yet).

If you would like to know more please PM, I don't want to hijack this thread

Peter

[Nemo121]

Hmm, the drop zone link isn't working and I can't seem to find this presentation everyone is talking about. Could anyone help me out with a URL please? Thanks.

[JeF]

Hi,

Wargamer.com is down for some time now.

Nevertheless, the FTP server seems to be up and running.

Direct URL for the presentation ZIP file is : ftp://ftp.wargamer.com/pub/DropZone/dep ... tiveAI.zip (22 MB)

I hope this helps,

JeF.