Neuroevolution Based Multi-Agent System with Ontology Based Template Creation for Micromanagement in Real-Time Strategy Games

I. Gabriel, V. Negru, D. Zaharie


This paper presents a multi-agent system that handles unit micromanagement using online machine learning in real time strategy games. We used rtNEAT algorithm in order to obtain customized neural network topologies, thus avoiding to complex network architecture. We use an ontology based template to create suitable input and outputs for unit agents enabling them to cooperate and form teams for their mutual benefit and eliminating communication overhead. The AI system was implemented using the JADE framework and the BWAPI handled communication between our system and the game. We have chosen Starcraft as a testbed. As a baseline we compared the in game AI as well as several other AI solutions that use adaptive mechanisms.



multi-agent systems; machine learning; real time games; neu-ral networks; neuroevolution; ontology

Full Text: PDF

Print ISSN: 1392-124X 
Online ISSN: 2335-884X