Rbot - The Intelligent Agent in Unreal Game Environment
Author
Benetis, Rimantas
Term
4. term
Education
Publication year
2002
Abstract
Dette speciale præsenterer Rbot, en kunstig modstander til videospillet Unreal Tournament. Rbot bruger bayesianske netværk, sandsynlighedsmodeller der vurderer, hvor sandsynlige hændelser er, til at træffe beslutninger under spillet. Agenten er implementeret i Java. Vi præsenterer to måder at opbygge indflydelsesdiagrammer på, en grafisk metode til at planlægge beslutninger under usikkerhed. Derefter præsenterer vi resultater for ydeevnen mod en indbygget Unreal Tournament-bot og en menneskelig spiller. Afslutningsvis drøfter vi, hvad resultaterne siger om anvendeligheden og ydeevnen af bayesianske netværk i et realtids-spilmiljø.
This thesis presents Rbot, an artificial opponent for the video game Unreal Tournament. Rbot uses Bayesian networks, probability models that estimate how likely events are, to decide what to do during play. The agent is implemented in Java. We present two ways to build influence diagrams, a graphical method for planning decisions under uncertainty. We then report performance results against a built-in Unreal Tournament bot and a human player. Finally, we discuss what these results say about the practicality and performance of Bayesian networks in a real-time game environment.
[This abstract was generated with the help of AI]
Documents