Although evolution has proven to be a powerful search method for discovering effective behavior for sequential decision-making problems, it seems unlikely that evolving for raw performance could result in behavior that is distinctly humanlike. This chapter demonstrates how humanlike behavior can be evolved by restricting a bot's actions in a way consistent with human limitations and predilections. This approach evolves good behavior, but assures that it is consistent with how humans behave. The approach is demonstrated in the UT^2 bot for the commercial first-person shooter videogame Unreal Tournament 2004. UT^2's humanlike qualities allowed it to take 2nd place in BotPrize 2010, a competition to develop humanlike bots for Unreal Tournament 2004. This chapter analyzes UT^2, explains how it achieved its current level of humanness, and discusses insights gained from the competition results that should lead to improved humanlike bot performance in future competitions and in videogames in general.

[Though this chapter was published later, an earlier paper describes how this bot was improved for later competitions: paper]

PDF

In Philip F. Hingston, editors, Believable Bots, 119--150, 2012. Springer Berlin Heidelberg.

@inbook{schrum:believablebots12, title={Humanlike Combat Behavior via Multiobjective Neuroevolution}, author={Jacob Schrum and Igor V. Karpov and Risto Miikkulainen}, booktitle={Believable Bots}, editor={Philip F. Hingston}, publisher={Springer Berlin Heidelberg}, pages={119--150}, url="http://nn.cs.utexas.edu/?schrum:believablebots12", year={2012} }