Many seem to think that AI needs to be able to think. The true definetion would suggest this but all it really needs to be able to do is follow set mathematical rules which can cover almost all basics. THAT IS BY NO MEANS EASY TO CREATE EITHER but would be much more satisfying to play agaisnt than a AI that cheats
RECON- easy, skirmishes(fast cheap units, followed by more substantial unit) for air, land and sea
EG, minimun strength unmaned recon, followed by a blackbird(even though it would probably be safe to send the blackbird first anyway)
It all just depends on different techs, different cpabilities and other differing aspects from scenerio to scenerio
The problem is how should the AI determine what unit to do what and how many units it should use and what should it follow up with, how should it respond.
All can be sorted out mathematically(not as good as human thinking because its inflexable) BUT much better than current games AI and would be more fun and challanging than a AI that cheats.
I might post the foundations I believe would form the basis of such in depth calculations.
But for now it will suffice to say the AI's calculations must be based on its ability to determine the relative strengths and weakness of all nations(including its own) and all factors such as build capacity, known units, reserved units and all the way to factors like economic vunerabilities and so on and so on.
I don't imagine to many of you would want to read formulas so I'll just say I'm stuck on one fundemental or essential factor for the begginings of a great AI which is as follows
BEFORE a war how does the AI know what units its enemies have including there reserves? MY SUGGESTION IS NOT A CHEAT but the best I can think of is the AI should know what the other regions start with(as a experienced player would) and then it should assume (based on the other regions building capacity) that there forces are growing by a certain amount. eg. Army build cap x4 the AI assumes the regions army grows by +16 mid quality units.
BUT thats all very general(also believe it or not it is brief) and depending on the response to this post I'll give more detail as required/requested but DEVS I'll give you it all once I'm satified with it and don't worry it'll like a AI tactics chart and as easy to read and focuss on whatever you wish as a good PDF file
So anyone have a good formula how AI should first evaluate for relative powers of all regions including itself
VERY BRIEF reasons why the AI doesn't need to cheat
Moderators: Balthagor, Legend, Moderators
-
- Brigadier Gen.
- Posts: 539
- Joined: Jun 01 2005
- Location: Sydney
-
- Warrant Officer
- Posts: 43
- Joined: Aug 10 2005
One of the main problems with these things, is that we're not able to judge easily what an AI can and can't do, we can dream up stuff we want it to do, but its still hard to say wether it actually can do it or not.
Simply put, the Ai must be able to allocate and send out units to take areas, today what happens is that it can have HUGE stocks of units, and send maybe 3-5 units tops to attack the enemy.
It might not even be able to differentiate between the diffrent units either.
So what it needs its, knowledge about the units its moving + knowledge about how many enemy units its facing and then respond to that.
Today, it doesn't even know how to defend, it only defends military bases, unless those are stuck wtih the capital cities, it won't defend any landmarks.
I fear you can't expect the AI to do advanced moves really, its not clever enough by far, and I think Battlegoat needs to realise this and work with what they got. AI has proven in many games to be very effective IF the rules it has to obey by are few, and its choices limited.
Simply put, the Ai must be able to allocate and send out units to take areas, today what happens is that it can have HUGE stocks of units, and send maybe 3-5 units tops to attack the enemy.
It might not even be able to differentiate between the diffrent units either.
So what it needs its, knowledge about the units its moving + knowledge about how many enemy units its facing and then respond to that.
Today, it doesn't even know how to defend, it only defends military bases, unless those are stuck wtih the capital cities, it won't defend any landmarks.
I fear you can't expect the AI to do advanced moves really, its not clever enough by far, and I think Battlegoat needs to realise this and work with what they got. AI has proven in many games to be very effective IF the rules it has to obey by are few, and its choices limited.
-
- Brigadier Gen.
- Posts: 539
- Joined: Jun 01 2005
- Location: Sydney
For the most part were in complete agreement in fact you reworded alot of what I said
MOST IMPORTANTLY TODAYS TECH ALLOWS AI to do incrediable things. CHESS is the example, the only problem is chess has simple rules(even though a complex game) whilst sr2010 has complex rules it requires alot of programming
BUT I DO BELIEVE THAT AS LONG AS YOU CAN COME UP WITH MATHEMATICAL WAYS TO DEAL WITH A SITUATION AN AI CAN THAN IMPLEMENT IT.
it just maths, alot of maths and the rest is up to the devs.
BUT AS FAR AS NOT KNOWING MORE ABOUT WHAT THE DEVS HAVE ALREADY AND/OR ARE PLANNING TO MAKE THE AI DO, I AGREE, I WOULD ALSO LIKE TO KNOW MORE
what I wrote: The problem is how should the AI determine what unit to do what and how many units it should use and what should it follow up with, how should it respond.
Now I agree the AI under uttilizes its units I've been saying that for the whole time and you could probably find many post of mine about this issue. The problem is if the AI just blatantly put its units forward than it would leave itself vunerable to air drops, shore invasions which could devaste its economy like taking out power plants and that would mean the AI would actually be EASIER to beat if it weren't dug in around bases and citieswhat you wrote:
Simply put, the Ai must be able to allocate and send out units to take areas, today what happens is that it can have HUGE stocks of units, and send maybe 3-5 units tops to attack the enemy.
It might not even be able to differentiate between the diffrent units either.
So what it needs its, knowledge about the units its moving + knowledge about how many enemy units its facing and then respond to that.
MOST IMPORTANTLY TODAYS TECH ALLOWS AI to do incrediable things. CHESS is the example, the only problem is chess has simple rules(even though a complex game) whilst sr2010 has complex rules it requires alot of programming
BUT I DO BELIEVE THAT AS LONG AS YOU CAN COME UP WITH MATHEMATICAL WAYS TO DEAL WITH A SITUATION AN AI CAN THAN IMPLEMENT IT.
it just maths, alot of maths and the rest is up to the devs.
BUT AS FAR AS NOT KNOWING MORE ABOUT WHAT THE DEVS HAVE ALREADY AND/OR ARE PLANNING TO MAKE THE AI DO, I AGREE, I WOULD ALSO LIKE TO KNOW MORE
- Sebastiaan
- Colonel
- Posts: 376
- Joined: Aug 29 2005
- Location: the Netherlands
- Balthagor
- Supreme Ruler
- Posts: 22106
- Joined: Jun 04 2002
- Human: Yes
- Location: BattleGoat Studios
CRAP! The dangers of Admin priveliges
I meant to hit quote, instead I hit edit...
What Sebastiaan's post said before I screwed it up (did I say yet?) was to suggest that the AI could be made open source if it was modular then anyone in the community could mod the AI and make their own changes.
Sorry again about screwing up your post...
I meant to hit quote, instead I hit edit...
What Sebastiaan's post said before I screwed it up (did I say yet?) was to suggest that the AI could be made open source if it was modular then anyone in the community could mod the AI and make their own changes.
But, it isn't...Sebastiaan wrote:This is made possible if the AI is modular and stored in a seperate DLL.
Sorry again about screwing up your post...
- bergsjaeger
- General
- Posts: 2240
- Joined: Apr 22 2005
- Location: Woods Bend, Alabama,USA