Designing virtual bots for optimizing strategy-game groups

In the last decade, a permanent increasing in the popularity of videogames has happened. As the cost of the computational power has decreased, graphic games more realistic have been developed. However, although everybody in this industry knew that an intelligent and believable behavior of bots (or, in general, non-playing characters) is an important key to make a game fun to play, the experts have not showed real interests in this issue until recently. In this paper, we propose how a good strategy for optimizing the behavior of a team of bots (with roles between members and communication skills between each other) in the “capture the flag game” domain, could be designed and analyzed using a combination between swarming optimization techniques and mathematical analysis based in Markov models in order to improve the standard strategies that videogames use. This domain presents a particular case of the “exploration vs. exploitation” dilemma, a paradox that appears in numerous situations where systems needs being adaptable and learnable at the same time and solutions of the dilemma are built evolving balances between parts. The environment used to test the proposed model will be the Unreal Tournament virtual world.