Robust Opponent Modeling in Real-Time Strategy Games using Bayesian Networks

Opponent modeling is a key challenge in the Real-Time Strategy (RTS) games since the environment in these games is adversarial and the player is not able to predict the future actions of his/her opponent. Moreover, the environment is partially observable due to the fog of war. In this paper, we propose an opponent model that is robust to the existing observation noise due to the fog of war. In order to cope with the existing uncertainty in these games, we design a Bayesian network whose parameters are learned from an unlabeled game-log dataset so it does not require a human expert‟s knowledge. We evaluate our model on StarCraft, which is considered as a unified test-bed in this domain. The model is compared with that proposed by Synnaeve and Bessiere. The experimental results on the recorded games of human players show that the proposed model is capable of predicting the opponent‟s future decisions more effectively. Using this model, it is possible to create an adaptive game intelligence algorithm applicable to RTS games, where the concept of build order (the order of building construction) exists.