HQ
I've been experimenting with adding some AI techniques to a two-player,
imperfect information, continuous motion game that I call HQ. The game is
written in C++ and runs on UNIX machines with X-windows. My first effort
was to see how well a program would play using a traditional
alpha-beta search.
I find that the program plays very well in tactical situations,
but is easily defeated using simple strategic moves. This was no big surprise.
I was pleased to find out that even a 5-ply search can be performed in a
reasonable length of time. For example, a 66MHz 486 searches 5-ply in about
two minutes, and an Ultrasparc or 200 MHz Pentium Pro will do the search
about 10 times faster.
Although the lion's share of the work has been on representation and search
issues, I have worked a little on an X-window interface that allows either
a person to play the machine, or the machine to play itself. Here is a
position from the middle of a game I played.
Here is an
mpeg movie
showing me (the blue player) against the computer (the
orange player). The computer is performing a three ply search.
The orange pieces are only visible when they are within sensor range
of one of the blue pieces.
Here is what really happened
during this game.