Chess game needs a high computational cost in order to be played correctly by a machine. Our proposal relies on the use of grid computing and genetic algorithms (GAs). On the one hand, grid computing offers us the potential for deeper game tree analysis. On the other hand, genetic algorithms can reduce significantly the computational cost of a brute force search, obtaining a good solution in a very lower execution time. Combining these two approaches (grid + EAs) we can obtain a very good chess player program. In this paper, we present our first steps in this line. A system that evaluates a chess play finding the best solution for the current chessboard and also evolving to achieve better results in future iterations. Starting from an algorithm with low computational cost, small solution search space, sequential execution and designed for accepting any kind of piece, this paper shows how the system evolves into a version with higher computational cost and achieving better and real-time results thanks to the use of grid computing. The results obtained with the different versions of our system are presented comparing them, and obtaining interesting conclusions.
Tipo de publicación: Oral