[Freeciv-Dev] Re: AI strategy

[Raimar Falke <hawk@xxxxxxxxxxxxxxxxxxxxxxx>]

On Sun, Dec 02, 2001 at 03:42:05PM +0100, Petr Baudis wrote:
> > > Another issue, what about observing AI? Tuning behaviour and
> > > debugging will be a lot harder w/o that.
> > 
> > This is one of the reason that these code parts will be available to
> > the human player as advisers. You can debug them easier.
> These code parts? Which ones you mean? CMA? That's good, nevertheless
> not enough. I can very hardly imagine advisor which will tell you
> with which unit attack which other unit and for which unit take which
> ferryboat to transport to which continent in order to start offensive
> against which player... 

One of the basics of the design of the (still non-existing) agents is
that you set _all_ parameters of the agent. The agent by itself is
dumb. You set the policy and the agent will go to work. So there may
be a ferryboat management agent. You set the boats you can spare and
the units which are allowed to use the boats. The FMA then will try to
work as best as possible if somebody tell the FMA what the destination
of the units are.

> > that tuning any AI is hard.
> I said harder, not hard :^). We will probably need to expand the current
> observer mode to support spying other connections ;-) (obviously crying
> appropriatelly everywhere in order not to be abusable).

IMHO also nice would be a gamelog which can easy parsed and contains
every important data. This has the advantage that is can be saved and
diffed.

        Raimar

-- 
 email: rf13@xxxxxxxxxxxxxxxxx
 "When C++ is your hammer, everything looks like a thumb."
    -- Steven M. Haflich