Cooperative and learning agents in the AgentGame environment

OData support
Eredics Péter
Department of Measurement and Information Systems

In my thesis, I have worked with the game named AgentGame. I first aimed to create an agent team that's competitive compared to the participants of the AgentGame contest arranged in the fall of 2010; and, while working on that, to get familiar with the aspects of the game and the environment Jason in which an agent team can be created.

After creating an effective team which could even survive amongst the conditions of the game's second version released in 2011, I started to work on the new challenges introduced with AgentGame 2.0.

The increased area of the game brought new possibilities to experiment with several forms of cooperation between the agents. I have implemented some forms of cooperation and inspected their efficiency. The puddles that appeared in the second version set back the agents' chances of survival, thus making the usage of learning necessary, in order to evade the puddles learned before. In aspects of learning puddles, I have implemented and inspected several solutions, differing in complexity and efficiency.


Please sign in to download the files of this thesis.