New Google AI Challenge: Planet Wars

http://aerique.blogspot.com/2010/09/new-google-ai-challenge-planet-wars.html

by mnemonik • 16 years ago

84 40 16 years ago

aerique.blogspot.com

pohl• 16 years ago

Is the structure of this game isomorphic to some other real-world problem they're interested in?

Elite• 16 years ago

Of course not Ender. It's just a game. Just a game.

endtime3• 16 years ago

It's being run by UW undergrads - I think it's just for fun.

phreeza• 16 years ago

I like the "Advanced Strategy Guide". As of 16:22 UTC all it says is "I got nothin"

http://ai-contest.com/advanced_strategy_guide.php

Cyndre• 16 years ago

Yet another fun challenge in the form of an AI challenge. I really enjoyed the tron ai contest and I am looking forward to this one.

Good luck Colonizing and coding.

petercooper3• 16 years ago

To skip the 3 line interstital blog post: http://ai-contest.com/

aerique3• 16 years ago

Yeah, I don't understand why that link wasn't posted.

My blog post was just meant to get it on Planet Lisp and make the Lispers aware.

petercooper3• 16 years ago

As someone who writes similar types of post (though usually a bit longer ;-)) I share your pain.

mnemonik• 16 years ago

Heh, my bad everyone :)

This blog came in my Google Reader feed from Planet Lisp and I was mostly excited about the possibility of a CL package. I just submitted with my Reader + HN greasemonkey script (http://github.com/fitzgen/reader-submit-to-hn) without a second thought...

zeteo2• 16 years ago

Is there any place where the game's rules are exactly given? I'm particularly interested in how the growth rate and battles work. The Python package that I've downloaded contains no such info.

what2• 16 years ago

http://ai-contest.com/problem_description.php

zeteo2• 16 years ago

That's the first place I've looked and it contains no clear information about the questions I'm asking.

A. "the number of ships on non-neutral planets increases according to the growth rate for that planet."

What does "according" mean - simple addition? And what happens to the neutral planets - no growth? In which case, how many fleets do the neutral planets start with?

b. "If the result is less than zero, then your bot gains control of the planet."

And how many losses have you incurred - exactly the number of defenders? If the defenders win, do they suffer losses equal to the number of attackers?

The reason I'm asking is that, if all the game does is simple addition and subtraction (no defense advantage, no Lanchester-type law for combat), then there probably exists a very simple optimal strategy, and it's more a problem of mathematics than AI.

troutwine2• 16 years ago

I rather wish their forum was open. I'm only receiving a blank grey window when I run ShowGame over a game's text.

Alas.

Edit: I suppose the outage was momentary; the forum is rather hopping.

Estragon2• 16 years ago

I downloaded the source from Google Code, and found the invocation in a .php file there. (Grep for "ShowGame"). After adjusting for the fact that I had the python starter kit, the invocation worked without problems.

troutwine2• 16 years ago

I'm not familiar with the Google Code site. Could your provide a link?

I've downloaded my kit from: http://ai-contest.com/starter_packages.php

Also, for what it's worth: http://ai-contest.com/forum/viewtopic.php?f=18&t=405&...

I suspect I'm having a goofy Java issue, but that's likely a personal bias.

NationalHQ2• 16 years ago

Where's the html5/javascript package? :\

tumult• 16 years ago

I don't think <video> is going to help you here.

j_baker• 16 years ago

I'm still holding out for the COBOL package.

_delirium2• 16 years ago

It's probably still too early on in these non-board-game-like game-AI competitions to do serious comparisons, but one thing that'd be cool to figure out in the future is how many components are resuable between these kinds of problems, versus requiring different strategy or hand tuning / hand feature extraction. For example, could similar bots do well at both Planet Wars and Starcraft (http://eis.ucsc.edu/StarCraftAICompetition)? If not, are there parts that can be shared (e.g. factor out high-level strategy versus low-level unit control)? Etc.

endtime3• 16 years ago

Relevant: http://cs227b.stanford.edu (General Game Playing)

GGPs can pretty much play any game with no continuous or unbounded variables, i.e. any game with a finite number of states. Popular examples include tic-tac-toe, checkers, and Connect 4; here's a larger list: http://code.google.com/p/ggp-base/source/browse/trunk/ggp-ba...

Planet Wars could be played by a GGP, assuming some bound on the number of ships that could be created, but standard approaches like UCT, which don't take into consideration similarity between states [1], would fail hard against a specialized player.

[1] For example, if UCT saw a good move in a given state, it wouldn't apply that knowledge to the adjacent state in which only one ship is in a different location.

_delirium2• 16 years ago

Ah yeah, I'm quite familiar with GGP. I'm very skeptical of its "generality", though, once you get away from anything that's vaguely similar to chess. For example, try writing Starcraft in GDL (even a discretized version); it's not impossible, but I sure wouldn't want to do it, and the result wouldn't be tractably usable by any GGP engines either. I'm working on a project currently that needs a logical representation of videogames, but I'm not using GDL, because it's just a pain in the ass for videogames (and I want an internal time axis also, not a state-transition system).

But mostly I was thinking on the playing side: I don't think any of the current GGP approaches can even come close to playing strategy-type video games, because they don't typically have any real notion of, as you mention, similarity between situations, or a decomposition into hierarchical or interrelated concerns. A Starcraft bot that treated Dragoon microcontrol as an equivalent-level problem to build order probably isn't going to do well. I think the robotics community might actually have a better starting point: architectures with interrelated components like SLAM, planning, motor control, etc., as opposed to trying to turn the "robotics problem" into one giant state space to be navigated by one giant algorithm. Trying to build an agent that plays two quite different strategy games well (Starcraft and Planet Wars) might be one way of driving the development of those kinds of decompositions.

NationalHQ2• 16 years ago

Does this teach me how to think with artificial intelligence?

Tichy• 16 years ago

If you choose to solve it with artificial intelligence, then yes.

javert• 16 years ago

Is it just me or is having the navigation on the right hand side of the page really unusual?

petercooper3• 16 years ago

It's not the "standard" but sidebars are common enough for it to work. Back in the 1998-2002 area, right hand navigation bars had a spurt of popularity because of usability findings in relation to the ease of moving between the scrollbar and the items. With the advent of scroll wheels, though, this hack became less significant and left handed navigation began to rule once again.

icegreentea• 16 years ago

Seems like the standard 'blog' (or at least blogspot) format.

Groxx• 16 years ago

>Use of multiple processes or threads is prohibited.

How about mimicking them? Can one perform in-process "context switches" to do the same thing?

If no: where do you draw the line? Must each second's decisions be deterministic from the board state?

If yes: what's the difference? Ease of programming seems a poor metric, and you could just nab something that's already written that does this for you, resulting in nothing but increased compilation times.

a1k0n• 16 years ago

The reason for this restriction is because your process and your opponent's run simultaneously on the same machine, so you only get one CPU. Doing your own userspace threading or whatever is fine.

aharrison• 16 years ago

Is anyone aware of an AI game that has a homeostatic goal? I am thinking along the lines of a SimCity style game where random negative events happen and our AI would have to account and route around these types of issues. It would be interesting to see a simulation of that nature as compared to the normal competitive game (e.g. chess, go, reversi.)

bluesnowmonkey• 16 years ago

Galcon itself used to be mod-able, with several people creating bots for it. Someone rolled them all together in a single mod allowing you to pit different AIs against each other, tournament style. The whole package was an excellent sandbox for teaching AI to budding programmers.

Sadly, that's all gone. Mods are gone, the modding forum is gone. Rather than an engrossing programming game, Galcon is now a mind-numbing clickfest. I am delighted to see Google picking up the concept.

andreyf• 16 years ago

Do you know who worked on it, and why it's gone? I might be interested in making a extensible HTML5 version of it...

New Google AI Challenge: Planet Wars

Related Stories