| The Othello Museum | |||||||||||||||||||||||||
|
AiboReversi
Play Reversi, one of the favorite board games around, against your Aibo! When starting Gamepack One, select the AiboReversi game and place your Aibo lying down in front of your AiboReversi game board. Then select the position you want to play using an easy code by pressing the Aibo paw switches, and confirm your move by pressing another sensor (sensor locations vary between Aibo models). Aibo will then think for a little while (usually less than 30 seconds), and use an easy code to tell you where Aibo's move will be. The easy to understand manual lists all the commands you need to go through the game, from correcting mistakes you make to passing a move, to having Aibo tell you all the pieces on the board in case you made a mistake. You can play AiboReversi on the game board provided in the game set, or use any Othello(tm) game board you already have. AiboReversi is written using a 1-ply alpha-beta search on all supported Aibo models, making Aibo look forward in the game as it plays. Uhm? It's hard to say that 1-ply "makes Aibo look forward"... it's also hard to define it "alpha-beta" and very easy to believe that 1-ply is performed in "less then 30 seconds" (maybe it was microseconds?). I'm also wondering what's the level of such a machine at 1-ply (I know, I know,don't expect too much from a dog...) And what's the need for Aibo to sit in front of the board, if it doesn't do anything to read it? Mmmhh... a real best seller, I guess.
On August 22 2008 I received an upset mail from the developer of AiboReversi, Christian Meunier, for this article. After a few kind explanations, he came to me with a very interesting story about how the production and development of the software had gone on. It's a pleasure to publish it here!
Reading your original reaction, I guess that descriptions for the layman ("makes it look forward") don't mix well with algorithmic descriptions ("1-ply alpha-beta" - where the reader should think about http://en.wikipedia.org/wiki/Ply_(game_theory) and  http://en.wikipedia.org/wiki/Alpha-beta_pruning ). I'll remember that... Doing the project was quite a challenge... over the years I kept a constant contact with Sony, who have always told me that if I make something that I give away for free (i.e. only free game downloads) they would let me use their software development tools and libraries. If I wanted to make something nicer (hiring graphic designers, printers to make a professionnal looking game boards - which would cost me real money hence the commercial part), they didn't have any objection but they didn't want to let me use any of their technology - and made it clear they didn't even see how that could be done. So I was on my own. The only language for Aibo that Sony would let me use ( http://www.ouroboros.org/rcode_tutorial_1v2.pdf ) was extremely slow and so awkward that even writing a tic-tac-toe game needed weeks of work. Still Sony was taken by surprise as they didn't expect anything more than "robot, walk around" to be done with that language. After one or two games developed that way, even simple ones, I had pretty much reached the limit of what would be doable legally with the robot... so instead of abandoning the project, I took a few months to write a C compiler for Aibo that I would be legally entitled to use. That way I was able to develop new (complex) games without having to use the awkard R-Code language, although my compiler still couldn't (legally) go around Sony's speed restriction (i.e. the compiler's output was still in R-Code). So that's how a Reversi game came to be, that takes 30
seconds to calculate a move, with no game move library and looks only one move
ahead... ;)
|
Or try this other translation:
The flipping clock. Please just click some ads to keep this site alive, it's a minimal effort for you! |
|||||||||||||||||||||||
 |
|||||||||||||||||||||||||
/5