A Survey of Opponent Modeling in Adversarial Domains

Opponent modeling is the ability to use prior knowledge and observations in order predict behavior of an opponent. This survey presents a comprehensive overview existing opponent techniques for adversarial domains, many which must address stochastic, continuous, or concurrent actions, sparse, partially observable payoff structures. We discuss all components systems, including feature extraction, learning algorithms, strategy abstractions. These discussions lead us propose new form analysis describing predicting evolution game states over time. then introduce framework that facilitates method comparison, analyze representative selection using proposed framework, highlight common trends among recently methods. Finally, we list several open problems future research directions inspired by AI on related other disciplines.