Game AI: micromanagement in Star Craft.

Abstract

Real-time strategy (RTS) games are both popular among players as well as among academics. A typical RTS gameplay can be divided into macromanagement and micromanagement. Other researchers have successfully applied Case-based reasoning (CBR) and case-based planning techniques in RTS games. While several of those approaches have beaten random pools of static scripted computer opponents, only a few approaches are focusing on the micromanagement aspect of RTS games. Traces Based Reasoning (TBR) is an enhancement to the conventional CBR methodology that adds the use of user-system interaction logs (traces). We present a CBR system that we convert into a TBR system that is used for micromanagement in RTS games. We explore various ways of micromanagement representations as well as case retrieval mechanisms suited for a system using the TBR paradigm. We test the system against a dynamic opponent in a micromanagement setting to present some of the difficulties involved in such an approach.