Re: [Freeciv-Dev] AI Diplomacy

[David Pfitzner <dwp@xxxxxxxxxxxxxx>]

"Ba'rdosi Zolta'n" <hck@xxxxxxxxxxxxxxxxxxxxx> wrote:

>  Can you inform me about the future plans of AI development,
> especially diplomacy?  I think the current state is very unfriendly,
> the only thing this ai can do now is the total annihilation of
> every living thing. (Including me sometimes ;)

We would like to improve this aspect, but, so far, no-one
has done (or indicated they are doing) the necessary work.

>  Who maintains the ai part now? how can i reach them?

I am not aware of anyone particularly working on the AI at present.
Anyone who wants to have a go is welcome to do so.

Personally, I would like to work toward making client-side AI's
more viable, so that people can more easily work on experimental
(client-side) AI's without having to mess with freeciv proper.

To this end, I am thinking of:

1. Have freeciv know about player names (aka usernames, knicks, 
etc) as well as Ruler names.  This is to help support:

2. Allow multiple client connections per Ruler/race.
(Probably: the first player to connect is the primary one,
who can then permit or restrict other players to connect
as that Ruler, as observers who cannot make changes.  The
primary player can then also give observers the full ability 
to make changes, if desired.)  This would allow people to 
watch & learn; or multiple players to cooperate in managing 
a large empire; or, more to the current point, give an easy 
way to monitor the actions of a client-side AI -- without 
having to put any graphical stuff into the AI code.

3. Have an option such that individual players can be set
as "turn-based".  Such players would act in turn after
or before all the "real-time" players.  This is so that
client-side AI's could be set as turn-based, and then don't 
have to worry about doing things in real-time, since I imagine 
a real-time capable AI would be much more difficult to code 
(eg, if the AI thinks about something, and then by the time 
it has decided, the situation changes).  Ideally, when the 
turn-based players are moving, real-time players would still 
be permitted to do certain actions, so long as those actions 
would not affect the view of the turn-based players.  
(Eg, change city production, tax rates; allowing changing 
workers would be nice, but maybe not feasible since it could 
affect work-placement choices for nearby enemy cities.)

Of course after all that, one still has the difficult task of
writing a capable client-side AI (eg, comparable to the current
AI), let alone implementing any advanced stuff like AI diplomacy.
So maybe the above is not such a great plan after all...

-- David