EVO-CBG - An Evolutionary System for Automatically Generating Character Behaviours for Game Environments

摘要

In this paper we discuss the need to extend the standard types of character behaviours found in game environments if we are to create new and more compelling gaming experiences. We propose using techniques from Evolutionary Algorithms and research from Game Design to create a system that can help game designers extend standard types of behaviours. This system automatically produces behaviours that are designed to optimize predefined parameters in the game environment. We outline experiments conducted using an implementation of this system to produce behaviours for the game Ms. Pac-Man, along with providing an overview of the results obtained. Finally we discuss these results and the potential they show for our system to help game designers not only create character behaviours, but also diagnose the effect that adding or removing certain mechanics will have on the overall gaming experience.