Abstract

In the world of Artificial Intelligence (AI) for video games, there exist many options for implementing Non-Player Character (NPC) behaviors. Choosing which methods and tools to use at the beginning of development for fast iterative prototyping is difficult. This thesis proposes a methodology for simultaneously using multiple graph- and reaction-based symbolic AI methods. Specifically, we will adjust decision trees, state machines, and behavior trees to work under a unified system. This method is then implemented in a node editor tool inside the Unity game engine with a complementary runtime framework for executing the combined graphs, called flows. The results are presented in a demo project, including example flows of potential use cases. Finally, the methodology and the tool are evaluated, and further research topics are suggested.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]