Complete.Org: Mailing Lists: Archives: freeciv-dev: March 2002:
[Freeciv-Dev] Re: [Patch][RFC] AI can fly
Home

[Freeciv-Dev] Re: [Patch][RFC] AI can fly

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: Petr Baudis <pasky@xxxxxxxxxxx>
Cc: Gregory Berkolaiko <Gregory.Berkolaiko@xxxxxxxxxxxx>, "Per I. Mathisen" <Per.Inge.Mathisen@xxxxxxxxxxx>, freeciv-dev@xxxxxxxxxxx
Subject: [Freeciv-Dev] Re: [Patch][RFC] AI can fly
From: Raahul Kumar <raahul_da_man@xxxxxxxxx>
Date: Sun, 3 Mar 2002 20:27:39 -0800 (PST)

--- Petr Baudis <pasky@xxxxxxxxxxx> wrote:
> Dear diary, on Sun, Mar 03, 2002 at 06:10:42AM CET, I got a letter,
> where Raahul Kumar <raahul_da_man@xxxxxxxxx> told me, that...
> > > In general, I am toying with the idea to have an infoset for AI, an
> advice 
> > > database of sorts, which would contain info like:
> > > 1. what to build to protect from unit_type1
> > > 2. what to build to attack a city defended by unittype2
> > > 3. what research to go for in such and such situation
> > 
> > I've been toying with the reserach idea. It can be split into a few
> sections
> 
> This we already have. We're increasing the desire to research certain tech
> everytime when we think it would be cool to have some unit/building which
> requires this tech.
> 
> > Wonders(techs needed for wonders)
> > 
> > Sort the wonders we want. Drop useless wonders like Marco Polo or UN, rate
> > highly the good wonders.
> 
> Huh.. useless? Oh well..
> 

Yes. If you're wasting time building UN while I'm building Hoover Dam or Seti,
guess which one of us wins ?

> > This might be invariant. Only a few wonders become pointless in certain
> > situations i.e Great library becomes less and less valuable with few
> players,
> > becomes 0 with 2 players, very little with 3.
> 
> See advdomestic.c:ai_eval_buildings().
> 
>       if (id == B_GREAT) {
>         values[id] =
>             (total_bulbs_required(pplayer) * (100 - game.freecost)) * t *
>             (game.nplayers - 2) / (game.nplayers * 100);
>       }
> 

Problem right there. I do not see a generic rate_wonder function, where all
the wonders are rated by the their effects. I.e affects production flag etc.


> for your Great Library. All wonders and buildings are considered there.
> Techdesires for security buildings are maybe also modified in
> advmilitary.c:assess_danger().

Cool.

> 
> > Military - this is obvious. The quickest path to mech inf and riflemen, the
> > quickest path to howitzer and stealth bombers and tanks.
> 
> See each fourth function in aiunit.c and advmilitary.c ;)).
> 

That's my problem. It should all be in advscience.c.

> > Production - Sanitation, aqueduct
> 
> advdomestic.c
> 
> > Transport - railroad.
> 
> Ok. To be honest, I've no idea how/where this is considered.

True. Maybe railroad, bridge building, and airport techs, as well as other
transport enabling techs should have their own handling.

> 
> > Overwhelming science lead - philosophy, republic, railroad, the wonder
> Darwins
> > Voyage, all the science wonders.
> 
> Ad wonders, advdomestic.c - for the rest, we rate high techs giving us many
> other techs available. 

The path I described has never been followed by any AI. I have never seen an AI
select it's highest trade city, build all the science wonders there, celebrate
every turn to pump it to a high level. The Super Science city would really help
the AI.




__________________________________________________
Do You Yahoo!?
Yahoo! Sports - sign up for Fantasy Baseball
http://sports.yahoo.com


[Prev in Thread] Current Thread [Next in Thread]