Developing game AI for the real‐time strategy game StarCraft

Research into intelligent agents designed to play Real-Time Strategy (RTS) games has been limited compared to the amount of research that has gone into designing agents for playing board games such as chess. This is largely due to the fact that developing an agent for RTS is more complicated. There has been some positive development in the last couple of years though with the introduction of the Brood War API, which allows researchers to hook their AI into the popular RTS game StarCraft. StarCraft provides an interesting platform for AI RTS research due to its large variety of units and balance between playable factions. We designed an AI agent capable of making informed decisions in StarCraft. The agent uses Reinforcement Learning to improve its performance over time. Empirical evaluation of the agent shows this type of an approach to be viable in RTS games.

Authors dc:creator

• Aleksandar Micić 1986-
• Davíð Arnarsson 1988-
• Vignir Jónsson 1988-

Contributors dc:contributor

• Háskólinn í Reykjavík

• Gervigreind
• Forritun
• Tölvunarfræði
• Artificial intelligence
• Programming (Electronic computers)

Language dc:language.iso

en

Handle dc:identifier.uri

http://hdl.handle.net/1946/9143

OAI identifier oai:identifier

oai:skemman.is:1946/9143