Perceived Challenge
Authors
Jensen, Dan ; Nissen, Kim Jung ; Justesen, Morten
Term
4. term
Education
Publication year
2010
Submitted on
2010-06-04
Pages
200
Abstract
Denne afhandling undersøger Behavior Trees (BT), en udbredt måde at styre spilfigurers beslutninger og handlinger via en trælignende struktur. Vi præsenterer og formaliserer BT og udvikler en metode til at måle, hvor varierede og hvor udfordrende forskellige BT'er er for spillere. BT-systemet SMARTS revideres for at øge ydeevne og stabilitet. Vi udvider det med et generelt perceptionssystem (så AI'en kan opfatte objekter og begivenheder i spillet) og en scheduler (der fordeler og styrer opgaver effektivt over tid). SMARTS integreres fuldt i den Qt-baserede Gluon-spilmotor. For at evaluere bidragene konstruerer vi spillet The Quantum Sea. Resultaterne viser, at målingerne er nyttige, og at den reviderede SMARTS-version med scheduler og perceptionssystem fungerer bedre end den tidligere. Det Qt-baserede Gluon-rammeværk er anvendeligt, men mangler aktuelt nogle funktioner.
This thesis examines Behavior Trees (BT), a common way to control game characters by arranging decisions and actions in a tree-like structure. We present and formalize BT, and develop a method to measure how diverse and how challenging different BTs are for players. We revise the BT system SMARTS to improve performance and stability, and extend it with a general perception system (so the AI can sense objects and events in the game) and a scheduler (to assign and manage tasks efficiently over time). We complete the integration of SMARTS into the Qt-based Gluon game engine. To evaluate these contributions, we build a game called The Quantum Sea. Our results show that the measurements are useful, and that the revised SMARTS with the scheduler and perception system performs better than the previous version. The Qt-based Gluon framework is useful, though it currently lacks some features.
[This abstract was generated with the help of AI]
Documents