Abstract: Agent is a core term in the field of artificial intelligence. In recent years, agent technology has been widely studied and applied in such fields as autonomous driving, robot system, e-commerce, sensor network, and intelligent games. With the increase of system complexity, the research focus on agent technology has been shifted from single agent to interactions between agents. In scenarios with multiple interactive agents, an important direction is to reason out other agents’ decisions and behaviors, which can be realized through the modeling of other agents involved in the interaction, that is, opponent modeling. Opponent modeling is conducive to reasoning, analyzing, and predicting other agents’ actions, targets, and beliefs, thus optimizing one’s decision-making. This paper mainly focused on the research on opponent modeling of agents, and introduced the opponent modeling technology in agent action prediction, preference prediction, belief prediction, and type prediction. In addition, their advantages and disadvantages were discussed, some current open problems were summarized, and the possible future research directions were presented. 

Key words: decision intelligence, opponent modeling, game theory, agent systems, AlphaGo