Abstrakt
Developing intelligent virtual agents (IVAs) is a great challenge for computer programmers. Lifelike virtual environments present various obstacles, especially on the lower AI level.
Navigating through 3D worlds is notoriously difficult to handle properly as well as quick and appropriate reactions to rapid changes of the environment. When teaching basics of IVA development at our university, we noticed that students find the complexity of virtual environments intimidating.
Although lectures on AI theory help, a substantial amount of hands-on experience is indispensable to gain proficiency. We thus searched for ways to start with very simple tasks and at the same time keep the students engaged and motivate them to experiment with AI development at home.
In this paper we report on two 3D virtual environments we developed on top of Unreal Tournament 2004 for the introductory classes of our course. The environments are inspired by children's games and are focused primarily on combining simple high-level decisions with navigation.
Tournaments of bots were held for both environments to conclude parts of the course. Evaluation over two years of the course shows that the environments helped students to focus on subparts of the IVA development and that the tournaments motivated the students to experiment with IVA behaviors outside the borders of the course.