Adaptive game AI for gomoku

Abstract

The field of game intelligence has seen an increase in player centric research. That is, machine learning techniques are employed in games with the objective of providing an entertaining and satisfying game experience for the human player. This paper proposes an adaptive game AI that can scale its level of difficulty according to the human player's level of capability for the game freestyle Gomoku. The proposed algorithm scales the level of difficulty during the game and between games based on how well the human player is performing such that it will not be too easy or too difficult. The adaptive game AI was sent out to 50 human respondents as feasibility. It was observed that the adaptive AI was able to successfully scale the level of difficulty to match that of the human player, and the human player found it enjoyable playing at a level similar to his/her own. ©2009 IEEE.