AAAI Publications, Twenty-Third International FLAIRS Conference

Font Size: 
An Effective and Efficient Real Time Strategy Agent
Kurt Weissgerber, Brett J. Borghetti, Gilbert L. Peterson

Last modified: 2010-05-06

Abstract


Real Time Strategy (RTS) games present a challenge for the field of artificial intelligence. The size of the state space often makes machine learning solutions intractable and as a result, developers turn towards hand-tailored solutions such as scripts. In this work, we present the Killer Bee Artificial Intelligence (KBAI) agent which uses a predictive nearest neighbor classifier to make decisions in real time. KBAI has a no-loss record against four hand tailored scripts in the RTS game Spring.

Full Text: PDF