MiniRTS - Strategy Identification using Bayesian Grid Models
Author
Dancker-Jensen, Mads
Term
4. term
Education
Publication year
2008
Abstract
Rapporten undersøger, om bayesianske gittermodeller kan bruges til ræsonnement i realtidsstrategispil (RTS). Først byggede vi et forenklet spilmiljø, MiniRTS, til kontrollerede tests. Bayesianske gittermodeller er en sandsynlighedsbaseret metode, der repræsenterer spillekortet som et gitter og løbende opdaterer forventninger om, hvad der er sandsynligt. Vi designede flere scenarier for at sammenligne forskellige modelvarianter og se, hvilke faktorer der kan forbedre ydeevnen. Rapporten beskriver implementeringen, forsøgsdesignet og resultaterne, vurderer styrker og begrænsninger og peger på mulige forbedringer og fremtidigt arbejde.
This report explores whether Bayesian Grid Models can support reasoning in real-time strategy (RTS) games. First, we built a simplified game environment, MiniRTS, to run controlled tests. Bayesian Grid Models are a probabilistic approach that represents the game map as a grid and continuously updates beliefs about what is likely to occur. We created several scenarios to compare different model variants and examine which factors might improve performance. The report details the implementation, experimental setup, and results, evaluates strengths and limitations, and outlines possible future improvements.
[This abstract was generated with the help of AI]
Documents