Complete.Org: Mailing Lists: Archives: freeciv-ai: November 2004:
[freeciv-ai] Re: [Freeciv-Dev] (PR#10694) AI Builds Doomed Ferries and P

[freeciv-ai] Re: [Freeciv-Dev] (PR#10694) AI Builds Doomed Ferries and P

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: badamson@xxxxxxxxxxx
Subject: [freeciv-ai] Re: [Freeciv-Dev] (PR#10694) AI Builds Doomed Ferries and Passengers
From: "Per I. Mathisen" <per@xxxxxxxxxxx>
Date: Mon, 8 Nov 2004 15:15:49 -0800
Reply-to: rt@xxxxxxxxxxx

<URL: >

On Mon, 8 Nov 2004, Benedict Adamson wrote:
> Per I. Mathisen wrote:
> > The AI need to be taught to attack more intelligently
> > (massive attacks).
> Would you say the problem was that unit building is chosen on a
> city-by-city basis? Is it possible to implement coordinated attacks
> without changing to a centralised planner? I guess a centralised planner
> would be a major change.

No, I don't think we strictly need a centralised planner, but we do need a
replacement for fstk, which is both an attack planner and a city-by-city
unit want function.

Greg has suggested a sort of statistical build manager for the AI in the
past, and this not a bad idea, combined with other input, to ensure that
we build a bit of everything. I have suggested a queue scheduler, and I
think I have the notes for this ... somewhere ... and I will try to put
those up on the wiki later on. But nothing is carved in stone yet.

> The find_something_to_kill function seems to make some attempt to
> coordinate attacks, using the reinforcements_cost_and_value and
> invasion_funct functions. Evidently these are not good enough, but do we
> know why?

Not really. It has long been the idea that fstk should die. However, the
bar to replacing fstk is really high, so fixing it somewhat in the interim
might be worthwhile.

I had a huge patch a while back which rewrote fstk to use pf, but I
dropped it as I and Greg decided fstk was essentially unsalvagable.

  - Per

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