Qualitative World Models for Soccer Robots

Until now world models in robotic soccer have been mainly quantitative in nature, consisting of fine-grained (numerical) estimates of player positions, ball trajectories, and the like. In contrast, the concepts used in human soccer are largely qualitative. Moving to qualitative world models also for robots has the advantage that it drastically reduces the space of possible game situations that need to be considered and, provided the concepts correspond to those in human soccer theory, it eases the task of agent specification for the designer. In this paper we propose qualitative representations using ideas from spatial cognition and employing Voronoi diagrams. We also discuss how reasoning with these representations is achieved within our underlying agent programming