Multi-Agent Mental-State Recognition and its Application to Air Combat Modeling

Recognizing the mental-state, i.e., the beliefs, desires, plans, and intentions, of other agents situated in the environment is an important part of intelligent activity. Doing this with limited resources and in a continuously changing environment, where agents are continuously changing their mind, is a challenging task. In this paper, we provide algorithms for performing reactive plan recognition and embed it within the framework of an agent’s mental-state. This results in a powerful model for mental-state recognition and integrated reactive plan execution and plan recognition. We then apply this in an adversarial domain air-combat modelling to enable pilots to infer the mental-state of their opponents and choose their own tactics accordingly. The entire approach is based on using plans as recipes and as mental-attitudes to guide and contrain the reasoning processes of agents.