[freeciv-ai] Re: Borg AI.

[Gregory Berkolaiko <Gregory.Berkolaiko@xxxxxxxxxxxx>]

On Tue, 2 Jul 2002, Vasco Alexandre Da Silva Costa wrote:

> On Mon, 1 Jul 2002, Gregory Berkolaiko wrote:
> > On Thu, 27 Jun 2002, Vasco Alexandre Da Silva Costa wrote:
> 
> > I would change the structure a bit, especially military.  HQs tied to
> > cities can handle defence and at most counter-attack.  But serious attack
> > should be handled one level higher, in Army.  Also there should be few
> 
> Actually we had the same idea :-) We just didn't implement those other
> agents due to lack of time.
> Our idea was to use the HQs strickly for city defense, and use a General
> agent for deciding attacks. This agent would use Field General agentes to
> coordinate each army.

That is more elaborate that what I thought about.  I am not 100% sure this 
intermediate General is needed, but it is something that's definitely 
worth considering.

> > Also, Borg happily leaves the cities undefended, so I saw and entire civ
> > taken out by few lucky enemy chariots.  I suspect it can be easily tuned,
> > though.
> 
> Heh, yes. It is quite dumb defending. For one it doesn't have a very good
> concept of which zones are safe or not. Anyone that plays Civ realizes you
> must be most concerned with defending the borders and the capital, the AI
> does not.

Unfortunately it's hard to model "spatial planning".  But in this case you
can estimate the strategic threat to your city by tracing all inverse
paths from it, allowed by your ZOC (this is another application of
pathfinding which would need a special BMC-function).  If you don't reach
any enemy cities or potential landing areas, you are safe.  Of course this
would be very rare, there is probably access to water somewhere...  At
this stage everything become difficult: what is "almost safe" and how do
yu quantify it?

G.