Similarity-based Opponent Modelling using Imperfect Domain Theories

This paper proposes a similarity-based approach for opponent modelling in multi-agent games. The classification accuracy is increased by adding derived attributes from imperfect domain theories to the similarity measure. The main contributions are to show how different forms of domain knowledge can be incorporated into similarity measures for opponent modelling, and to show that the situation space of the opponent modelling approach is not required to be the same as the situation space of the opponent players. Our approach has been implemented and evaluated in the domain of simulated soccer.