Slashdot is powered by your submissions, so send in your scoop


Forgot your password?
AI Java Games News

AI Takes On Pac-Man 113

mikejuk writes "AI takes on Pac-Man — well, in fact it plays both sides. An annual competition challenges participants to write Java programs to control Pac-Man or the ghosts. It might not be chess, but it pits machine against machine, with algorithms going head-to-head as the AI ghosts try and eat AI Pac-Man."
This discussion has been archived. No new comments can be posted.

AI Takes On Pac-Man

Comments Filter:
  • Already Over (Score:3, Informative)

    by ctrimm ( 1955430 ) on Monday June 13, 2011 @08:57AM (#36424180)

    The contest for this year is now over.

    It would have been nice to know when it started... Guess I'll just have to prepare for next year.

  • by i.r.id10t ( 595143 ) on Monday June 13, 2011 @09:52AM (#36424614)

    To be fair, AI isn't required to beat the original quarter sucking machine - simply memorize the map patterns. Heck I remember buying a book that had the patters for all of the levels when I was a young lad...

  • by davidbrit2 ( 775091 ) on Monday June 13, 2011 @10:42AM (#36424988) Homepage

    Assuming they're pitting their Pac-Man bots against the logic of the original arcade game, there isn't a whole lot of complexity to deal with. All four of the ghosts behave in different, but very predictable fashion. In a nutshell, every ghost chooses its target space on the board differently, but they all close in on their respective target space in exactly the same manner. Knowing how these target spaces are chosen, one could probably write a reasonably effective AI player only needing to "think" a few hundred frames in advance. This is difficult for a human player to do in real time, of course

    There are a couple of interesting articles about how the game works. If you've played the game and know how unpredictable the ghosts can seem at times, it's remarkable to find out that the algorithms behind their behavior are so incredibly simple. I used to wonder if the game employed some kind of sophisticated path-finding algorithm like A*, but it's actually nowhere near that level of complexity.

    Anyway, this seems like it would make a cool undergrad project for an AI class.

"How many teamsters does it take to screw in a light bulb?" "FIFTEEN!! YOU GOT A PROBLEM WITH THAT?"