Exploring Applications of the “NPC Engine” MCTS Planner

Abstract

This bachelor thesis addresses the development of a game with the purpose of demonstrating the NPC Engine (NPCE), a Monte Carlo Tree Search (MCTS) planner developed at the Game Technology Center (GTC) of ETH Zu ̈rich, contributing to research in emergent narrative. The document outlines the step-by-step design process, exploring five initial ideas and elaborating over the chosen concept. The final product, built using the Unity game-engine in conjunction with the NPCE, is then presented, along with its limitations. We then present the outcomes of a quantitative user study conducted to evaluate the demo. The findings suggest that it offers an enjoyable experience but requires further refinement to comprehensively showcase all NPCE functionalities. Lastly, we make proposals on how to improve the NPCE workflow based on our own experience working with it, and we elaborate on the problems that future work should address in order to improve the demo.