iThink: A Library for Classical Planning in Video-Games

摘要

Academic artificial intelligence (AI) techniques have recently started to play a more central role in the development of commercial video games. In particular, classical planning methods for specifying a goal-oriented behavior have proven to be useful to game developers in an increasingly number of cases. Motivated by the fact that there is no clear standard for developing a goal-oriented behavior in video games, we present iThink, a framework that allows the use of academic techniques for classical planning in order to achieve goal-oriented behavior in a real game developing environment. In our work we focus on STRIPS, a well-studied framework for classical planning, and Unity3D, a popular game engine that is becoming an emerging standard for, so-called, "indie" game development. Except for being a useful tool for game developers, we believe that iThink can be used in education, providing a modern and fun environment for learning and experimenting with classical planning.