Principal Investigators: Zev Battad
Modern open-world videogames are large virtual environments, providing increasingly realistic simulations of the real world. A single game could have hundreds of locations, characters, and items modeled, giving players the freedom to immerse themselves in the complex system of interactions represented by the virtual world. However, as intricate as open-world videogames are, they also require a deep understanding of the game's complexities for players to be able to experience them at their fullest. This represents a barrier of entry, as players wishing to experience the true depth of a game must take a long time to learn its intricacies. Players turn to sources of information outside the game to alleviate this, at the cost of breaking immersion in the game world. Virtual characters that are able to navigate the complexities of large, open-world videogames in a way similar to players may provide an alternative source of information, all the while acting and appearing more immersive.
In this project, we present a semantically enabled method for game AI to plan over the entities in large, virtual environments in order to perform tasks in the environment. The method is brought to a specific virtual environment - the video game The Elder Scrolls V: Skyrim by Bethesda Softworks - to accomplish a specific task - acquiring entities called Items in the game. The method generates plans to obtain items in the game, including game locations to travel to, characters to interact with, actions to take, and items to obtain. Plans are actionable by game AI in Skyrim, allowing AI-backed characters to act more intelligent and immersive.
The method is supported by an OWL ontology, encoding the main superclasses of entities in Skyrim alongside game actions relevant to obtaining Items. The ontology connects the large, and disparate, population of entities in Skyrim under a common reasoning scheme. SPARQL queries are shown which extract plans from the ontology.
This ontology was developed during the Ontology Engineering class at RPI during the Spring semester of 2016.
Please visit the course page to find out more.