research: question to vic (Full Version)

All Forums >> [New Releases from Matrix Games] >> Advanced Tactics Series >> Mods and Scenarios



Message

kafka -> research: question to vic (12/1/2007 4:28:01 PM)

I plan to implement a component bases nation specific research tree, so you do not simply research a complete tank but a just a design and have to research some other components in order to be able to produce the item type, i.e.

you want to produce pzkpw Ia, so you have to research the pzkpfw I design before and additionally the following components: MG 13, the Krupp M305 engine and the ZF Aphon FG 35 transmission (just an example). So my question is: Would the AI be able to deal with the relative complexity of such approach?



Vic -> RE: research: question to vic (12/1/2007 6:02:44 PM)

i doubt it.

and though it is a bit of strange the AI actually is using the statistics of the SFType referred to in the picture of the research field to determine if it wants to buy that research field.

kind regards,
vic



kafka -> RE: research: question to vic (12/1/2007 6:08:13 PM)

ok thanks vic

so this means the AI expects each research field resulting in an item type (=sftype) it can produce?



seille -> RE: research: question to vic (12/1/2007 6:14:18 PM)

@kafka
And you think i want to research several parts just to be able to produce a single tank ?
I would need a half hour for research only and had to use a calculator the whole time [;)]
So why these complexity ? The AI can never handle that.



Vic -> RE: research: question to vic (12/1/2007 6:20:57 PM)

yes it expects it. but you could fool it.

the thing is that the AI tries to buy the researchfield that it thinks will give it the ability to produce SFType that is most usefull for it at the moment.

If it think Light Tank II will be most needed and you put that picture in the Engine Research field is will resarch the Engine Research Field. If you also put it in the 50mm gun researchfield it will try to buy that field next turn.

I think it should work. Just make sure that most usefull combo's have an SFType picture that is interesting for the AI to produce.

Work around is to write events for the AI to give it research for free every X rounds.

kind regards,
Vic



kafka -> RE: research: question to vic (12/1/2007 6:59:34 PM)

thanks vic

I do not want to fool the AI but just to find out the modding capabilities and constraints of the game

@seille

the game provides the feature of defining up to 5 research fields for an item type, I just wanted to clarify how this works, whether it is possible to define a flexible research tree where you can select between different research paths - yet I can implement only a structure based on what the AI can handle



jakobrogert -> RE: research: question to vic (12/9/2007 2:23:22 AM)

In my super koopa wars mods, I'm testing the AI's ability to handle the mod features and most basic things seems to go down well (Infantry production, transport units, engineers). But the A.I has yet to research something in a single game I've played against it. I'm testing on a 20x20 (very small) map but recall seeing the AI do research even on small maps, when playing the standard game.

So, what I'm wondering is what I could possibly have done super-wrong in order to prevent the AI from spending it's research PP's. I have adjusted the AI's attitude towards saving PPs for research to 0.2 (20% as I understood Vic's tutorial), but that hasn't helped.
In the AI logs, I can see the AI prioritizing Researches/technologies, but fails utterly to deploy a single Unit, except those I give it from the start.

Would be happy for any pointers.

Best Regards
Jakob



Vic -> RE: research: question to vic (12/9/2007 2:25:16 AM)

Did you set AI Role scores for the different SFTypes? Did you try making the AIRole scores of the upgrades cq research higher then the others?



jakobrogert -> RE: research: question to vic (12/9/2007 2:35:04 AM)

Yes, everyone has an AI role score, but I haven't tried making the units further down the tech tree score higher. My tech-tree isn't so much about upgrades to existing Units as new Units with new functionality, but I guess the AI sort of needs it straight to it's face regarding appropriateness of Units.
Thanks for the quick reply!



jakobrogert -> RE: research: question to vic (12/9/2007 6:16:25 AM)

Another AI-question (since I still can't seem to make the AI research):
Does the AI compare the price of research, to for example, new Units? In my Mod, all Units are a lot cheaper compared to their price in standard Advanced Tactics. Since I didn't change the price (in production) of political points, might the AI simply disregard teching because of the price, in favor of trying to just overrun me with whatever it has?



Bombur -> RE: research: question to vic (12/9/2007 11:59:15 AM)

-Maybe you could simply forget the AI. A mod like that could be designed only for PBEM play???



Vic -> RE: research: question to vic (12/9/2007 11:59:28 AM)

hi,

first of all.. why dont you attach an AI log to this thread.. i'll take a look.

next do all the researchfields have a SFType graphic linked to them? the SFType number specified for the graphic is the thing the AI will think it will be able to build if it researches it.

it should pick a research field for an SFType that can upgrade most AND that will be best in battle against existing enemy troops.

kind reg,
vic



jakobrogert -> RE: research: question to vic (12/9/2007 8:02:06 PM)

Hi again and thank you for your persistence in helping out. =)


Seems tweaking enough finally made the AI purchase a technology (after increasing the value of how much it should save for research and making the start infantry the infantry role score of 40 and the elite infantry the AI bought infantry score of 70).
The technology in question gives elite infantry but the AI never built it, but I'm starting to realize it's about tweaking. I guess the AI is reluctant to switch and field new Units unless there's a very direct usefulness for the Unit (it fields Armored Cars in standard AT quite quickly, but those provides mobility and great anti-infantry power, which is why I guess it almost always goes for the armored car).
However, my Units doesn't perhaps have the steep difference that an armored car and a bunch of infantry Units has. Also, if I may guess, I'd say the AI isn't researching at all when it's under pressure to get the amount of Units it wants; political points for Units are more important to it than research. In the test scenario, the first technology purchase the AI did was a couple of rounds into the game, when it had quite a large amount of Units.

Since the AI actually bought a tech, I'm on the right way, why I'm not posting an AI-log here.



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.7659912