With modern video games frequently featuring sophisticatedand realistic environments, the needfor smart and comprehensive agents that understandthe various aspects of complex environmentsis pressing. Since video game AI is oftenspecifically designed for each game, video game AItools currently focus on allowing video game developersto quickly and efficiently create specificAI. One issue with this approach is that it does notefficiently exploit the numerous similarities thatexist between video games not only of the samegenre, but of different genres too, resulting in adifficulty to handle the many aspects of a complexenvironment independently for each video game. Inspired by the human ability to detect analogiesbetween games and apply similar behavior on aconceptual level, this paper suggests an approachbased on the use of a unified conceptual frameworkto enable the development of conceptual AIwhich relies on conceptual views and actions todefine basic yet reasonable and robust behavior. The approach is illustrated using two video games,RavenandStarCraft: Brood War.
展开▼
