Analyze and guess type of piece in the computer game intelligent system

Abstract

Siguo game is an interesting test-bed for artificial intelligent research. It is a game of imperfect information, where completing players must deal with possible knowledge, risk assessment, and possible deception and leaguing players have to deal with cooperation and information signal transmission. Since Siguo game is imperfect information game that the player doesn't know the type of piece and strategy that opponent moves, to exactly guess type of opponent' piece is a very important parameter to evaluate the capability of Siguo game program. In this paper, we first construct a fuzzy type table by analyzing more than one thousand different embattle lineups (i.e. chess manuals) of Siguo game, and then we present a algorithm that updates type table by using information from opponent during playing game. The updating type of pieces algorithm is designed by considering the two strategies, i.e. optimism and pessimism based on the fuzzy notion. At last we give a method to guess the type of piece by using fuzzy type proximity relation between two neighboring pieces. © Springer-Verlag Berlin Heidelberg 2005.