Ant Colony Optimization in games?

[BULL]

Archived from groups: comp.ai.games (More info?)

Does anybody know any papers (preferably downloadable) that discuss
how ACO might be used in game playing? e.g. othello, 5-in-a-row,
backgammon, tic-tac-toe, etc.

 

[Chris]

Archived from groups: comp.ai.games (More info?)

I did some work on this when I was a student and gave a paper at
Metaheuristics International in 2001. I was looking at cooperative and
antagonistic actions in a simple maze, involving agents that moved
around a simple grid-maze trying to point at each other without being
pointed at. The algorithm got each agent to simultaneously develop a
path for itself alongside the expected path for the opponents so that
it was able (to a limited degree) to predict the opponents' actions.

The study was a bit naive and I have moved away from the subject since
then. I can email you the extended abstract if you like.

 

[Guest]

Archived from groups: comp.ai.games (More info?)

Chris wrote:
> I did some work on this when I was a student and gave a paper at
> Metaheuristics International in 2001. I was looking at cooperative and
> antagonistic actions in a simple maze, involving agents that moved
> around a simple grid-maze trying to point at each other without being
> pointed at. The algorithm got each agent to simultaneously develop a
> path for itself alongside the expected path for the opponents so that
> it was able (to a limited degree) to predict the opponents' actions.
>
> The study was a bit naive and I have moved away from the subject since
> then. I can email you the extended abstract if you like.
>

Hi. I'm very interested in it. Could you email it to me, too?

Regards,
sakurai.

 

[BULL]

Archived from groups: comp.ai.games (More info?)

> Chris wrote:
> > I did some work on this when I was a student and gave a paper at
> > Metaheuristics International in 2001. I was looking at cooperative and
> > antagonistic actions in a simple maze, involving agents that moved
> > around a simple grid-maze trying to point at each other without being
> > pointed at. The algorithm got each agent to simultaneously develop a
> > path for itself alongside the expected path for the opponents so that
> > it was able (to a limited degree) to predict the opponents' actions.
> >
> > The study was a bit naive and I have moved away from the subject since
> > then. I can email you the extended abstract if you like.
> >
>
> Hi. I'm very interested in it. Could you email it to me, too?
>
> Regards,
> sakurai.

I was the original requestor, but didn't receive anything. Just in case you
lost
my email address, could you please send your paper to:
bullFISHinternode.on.net (replace FISH with an @)

Cheers,
Bull.

 

[Chris]

Archived from groups: comp.ai.games (More info?)

Sorry - I have been away from the office over Christmas and took a
while to catch up. I have emailed the extended abstract to you. Make of
it what you will. Ask if you want to know more. I'd be interesting in
hearing if you make anything useful of it, or if you manage to get ACO
to work in a game using any other method. Since I was using maze games,
the ant metaphor made lots of sense. I'm not sure how well it will work
for the classical games you mention.

Chris

 

[BULL]

Archived from groups: comp.ai.games (More info?)

> Does anybody know any papers (preferably downloadable) that discuss
> how ACO might be used in game playing? e.g. othello, 5-in-a-row,
> backgammon, tic-tac-toe, etc.

I found an interesting paper titled "When Ants Play Chess":
http://www-poleia.lip6.fr/~drogoul/papers/Drogoul.Maamaw93.pdf