My major research activity currently involves the study of game playing programming. There are many games that are interesting from a programming and Artificial Intelligence (A.I.) point of view. I am currently programming two such games. One is named Invader and plays the game Amazons, a fairly new game that is attracting interest in the game programming community because it is not as difficult as go, but provides interesting challenges not found in many other games, like chess or checkers. The other is called Wanderer and plays the game of Havannah. Havannah is interesting for most of the same reasons as Amazons. I am seeking graduate and undergraduate students to work on projects related to these two games.
One of my main interests is to expand the MCTS (Monte-Carlo tree search) algorithms that exist in both programs. In the case of Wanderer we already have a basic MCTS engine in place but efforts to improve it have proved elusive. Though "heavy playouts" have proven successful in go programming we have not managed to have similar successes in Havannah. I suspect part of the reason is that we simply have not dug deep enough into Havannah and have not programmed in the proper data structures that will allow us to quickly find the more promising moves or conversely quickly eliminate the unpromising moves.
Things are a bit more interesting (and complicated) in the case of
Amazons. Invader does not use a pure MCTS approach. Instead, during the
random playouts rather than simulating to the end of the game it stops after
only 6 or 7 moves and then uses an evaluation function to determined the
status of the position. Not much research has been done on this approach and
so there are many opportunities here for interesting graduate projects.
Send me email: firstname.lastname@example.org. Last updated 08/23/2012.