Single-Agent and Adversary Search In Combinatorial Spaces: Fun With Puzzles and Games

Bart Massey
Department of Computer Science, Portland State University

Abstract: The essence of the difficulty of puzzles and traditional games lies in their requirement to choose wisely from a large set of possible futures. These choices are made by manipulating combinations of a small set of elements. Most modern computer programs that successfully tackle problems of "intelligent" puzzle-solving or game play do so by shallowly examining the consequences of a wide range of choices at each decision point.

This talk explains some theoretical underpinnings of this "combinatorial search" problem and briefly surveys the methods used for combinatorial search in puzzles and games. A couple of examples from past and current research will be discussed: specifically optimal Yahtzee and optimal Boggle boards. Finally, some consideration will be given to outstanding problems in this area.

Bart Massey ('87, Physics) is an Assistant Professor of Computer Science at Portland State University where his research focuses on open source software engineering, and on software applications of combinatorial search and other AI techniques. His doctorate is from the University of Oregon Computational Intelligence Research Laboratory (CIRL), with a dissertation on search direction in general-purpose planning.