MacGyver Virtual Agents: Using Ontologies and Hierarchies for Resourceful Virtual Human Decision-Making

John T. Balint and Jan M. Allbeck

Extended Abstract. Proceedings of Autonomous Agents and Multi-Agent Systems, 2013, pages 1153-1154.


Virtual Humans are becoming an ever more important asset in games, movies, and training simulations. The ability for virtual humans to select appropriate actions and objects for plausible behaviors is vital to creating believable and resourceful agents. For this purpose, a decision making methodology using hierarchies of ontologies for both actions and objects is presented. Objects are given semantic information such as affordances and physical properties. Affordance theory is applied to determine viable candidate objects for behaviors, and agents learn which objects are better choices than others.

Paper (pdf)