AMAI 3.0 Plans - Dymanic real time AI

[Strategy Master]

This next year should be an interesting one. What i want to plan is an AMAI 3.0 featuring fully real time strategies. Scouting its enemy will detect the units it comes accross and adapt to building those sorts of units. AMAI will start to become much more dynamic. Modders will only need to define the units and there usages as well as the rules etc then the dynamic systems take over and control everything.

Whats this mean for strategies themselves. Well there will be no actual strategies par say, just profiles. But your have different strategy agendas instead. So maybe have a rushing agenda which focuses on attacking you early, keeping the pressure on at tier 1 and harrassing. Another will focus on defense and expansion and then pick you off slowly with harassing and maybe unexpected harrassments (invibility anyone), and another will just tech aiming to build up alot of tier 3 to kill you off. AMAI will pretty much stay with there chosen agenda all game.

Of course it will respond to what it sees when scouting so sees a army of griffons its gonna go ahh need anti air fast and build those although it does using different ways depending on the agendas. It will realize you have a mine only once its seen it.

In words its much more effiecient macro. Also on the way i can hopefully improve team play considerably. AMAI 2.52 already has pretty much exellent micro systems so I want to spend more times on other aspects of AMAI.
THe result is further human like opponents and hopefully challenging where every game of AMAI could be unexpected.

Of course this is only the planning and i have already ideas of how to implement. Only bad news is this is potentially my last year of uni meaning i may not have much spare time to work on amai due to expected masses of work and this increasing (playing games withdrawl symptoms due to constant modding :P). I'm hoping i can make some headway to completion in my spare time.

[moyack]

That's for me the perfection!!! I wish to see the results.

[lord999]

thanks for your effort Strategy Master!

[Av3n]

This sounds good

-Av3n

[Vexorian]

sounds great!

[archpsi]

I would like to help if its possible :), but not for at least the next 2 months.
anyway what algorithms will you be using or is it mainly just script driven?

also with the upcoming release of sc2(might be a year, but im sure the beta will begin long before release), it might be time to think of SC2MAI :) or something.

[[nVp]Zitrax]

Greetings Strategy Master,
I'm Zitrax from No Virgins Please.

When i was still learning to play Warcraft 3 TFT,
and i was pretty weak, i use your AMAI at normal level.
I felt like i play with a real guy in battle net.Anyway,
im pretty much in love with AMAI so i cannot leave it.
AMAI open my doors, AMAI computers was my first
training partners and i wont forget that.After i felt like they
are too weak i begin my real training at battle net.There was no
real difference, i felt like playing with AMAI.So i quckly advance
in the game learn more and more until my postion now, a Shaman
in No Virgins Please.I just wanna thank you for all you efforts for all
you hard work to make it better and better, for all your long hours
infront of the computer, for all you free time spend on the PC.I wanna
say THANK YOU!

Second when i heard about AMAI 3.0 i was pretty confused, i thought
that AMAI project was long time forggten and no one update AMAI.
Good luck with the new AMAI, i hope its even better than the
last ones.Also dont spend too much time infornt the computer, u probably
study so study hard, but when u have free time work a bit on AMAI.
I hope you will make it soon.

Best Wishes!
[nVp]Zitrax

[Strategy Master]

@archpsi: Thanks for the offer. I havn't set up concrete plans yet of what and how i will be doing things just yet. As i said those are just the basic plans to enact. My goal will be to get the real time strategy system working first as its the hardest part to envision. So it will first just dynamically choose what to do by cheating to see the enemy etc. Then we work on the other features like strategy agendas and AI scouting.

@Zitrax: Cheers its good to here how AMAI helped train people for battle.net. No was never forgotten, AMAI 2.52 is pretty much everything originally in AMAI now working as was originally intended which took pretty much 2-3 years i guess now. This should make work onto AMAI 3.0 be really easy as lots of the nitty gritty stuff is out of the way and its actual mechanics of stuff that are now being changed. I actually could be studying alot harder. I'd probrably be then on track to get first instead of 2.1 but i'd hate to dull the mind to get there. That destroys the potential for any authentic creativity. (Guess the film quote if you must :P)

[[nVp]Zitrax]

Oh, yeah and one more thing,
Your probably notice my name is "Zitrax" like one of your AMAI computers, i chose it in honor of your great work on AMAI and its sounds pretty good too :-)

Can you give us more details about your future project AMAI 3.0 next year.I am pretty exited already, i played so much game with AMAI when i was newbie and now i dont have patience for playing again with 3.0 :-)

EDIT: I just downlaod AMAI and for all my fans (lol) a quick ownege of AMAI (Insane)

[Zalamander]

Hehe I just sat and picket lots of classic names of old computer games and scifi series mostly when I wrote down the AMAI names fast to get them over with.

Zitrax is a pretty cool old shoot'em up game I had on my Amiga 500. :)
http://www.lemonamiga.com/?mainurl=h...hp%3Fid%3D2510

[deutschepharma]

This is great news indeed, I'm glad you didn't get tired of this excellent project!

I would be glad to help in a constructive kind of way, in any way I can - if you put some suggestions/discussion thread, i promise to visit frequently and participate.

[Goose]

Awesome! Warcraft3 is still my favourite game. I don't play it as much anymore, and certainly not as obsessively as I used to, but I still have it installed and I come back every 6 weeks or so for a game vs AMAI.

Really glad to hear you're still working on it! And great to see people sounding off in this thread that others are still out there appreciating all the hard work you put in.

Thanks a million!

[Quitch]

I'm excited about the prospect of improved teamplay, as this is quite the chink in the AMAI arsenal IMO. Here's hoping you can see it through, or perhaps these plans will come to fruition in Starcraft II?

[thenonhacker]

Thanks SM! w00t!!!

Hey, when Starcraft 2 comes out, please let your AMAI Framework benefit it!
In fact, I guess it will be 2x Exciting and 2x Richer (the World Editor is very improved!!!)

[Strategy Master]

I'm excited too about starcraft 2, in all fairnest i'm only excited by the editor capabilities being i never really played the original starcraft :P No doubt its a superb game though

Anyway more importantly apologies i been rather missing of late. Basically really been heavy in work load this year. 4 weeks doing 4 courseworks and now i got like 2 days to revise for exams. Who ever made these plans in curriculum timetabling i dont know. AMAI could so reactify that :P I dont know how many of you know but i've also been a little on the oblivion modding scene so that also taking up some of my spare time i had.

But I come to talk to you about units. Its the first time really i thought about the nitty gritty workings of the real time dynamic unit systems. But the idea is to have a percentage system(or if possible maybe extend to some sort of bayesian network system, But for starters lets stick with the simple idea below).

We sum up enemy strength in different categories and basically percenticise it. So lets say enemy has 20 strength heavy and 5 strength air. Thats 20/25 percent chance AMAI responds by building a unit to counter heavy.

Next it does the same this time with all the units it could choose to counter the heavy. So lets say for example purposes we have griffons given a value of 12 and footmen given a value of 15.

That means its a 15/27 percent chance to build a footmen. Of course realistically there will be more than just two measily units involed.

The idea though is because strategies wont really exist anymore we can just use the entries to make like percentage bonuses for customizability sake.

So ModPercent(FOOTMAN, Heavy, 5) will add a 5 base chance against heavy armor for footman. This keeps an aspect for modders to use and personlise the real time system.

This same system can extend then to upgrades. Each unit built increases the percentage of wanting to get this upgrade.

Heroes will have to be some sort of special case in how and when they are trained.

If done properly it will work wonders as if enemy is spamming the same type the percentage chance of building certain types of units to counter them will greatly increase and hopefully reach the 99% mark as the other unit percents decline.

If you have any ideas or suggestions on the idea above feel free. Hopefully i can start something after exams as start of semester 2 will be at least a little quieter again. If i can get this part working ill update the Tech build with it demonstratingly working (no pre fixed strategies woot)