SPP RoboCup in Bremen - Summary

Our main research interest is the automatic diagnosis of the strategies of other agents, in particular, of the opposing team in RoboCup. A new challenge in the development of autonomous physical agents is to model the environment in an adequate way. In this context, modeling of other active entities is of crucial importance. It has to be examined, how actions of other mobile agents can be identified and classified. Their behavior patterns and tactics should be detected from generic actions and action sequences. From these patterns, future actions should be predicted, and thus it is possible to select adequate reactions to the activities of the opponents. Within this scenario, the other physical agents should not to be regarded individually. Rather it should be assumed that they form a self-organizing group with a common goal, which contradicts the agent’s own target. In consequence, an action of the group of other agents is also a threat against the own plans, goals, and preferred actions, and must be considered accordingly. Acting under the consideration of the actions of others presupposes a high degree of adaptability and the capability to learn from previous situations. Thus these research areas should also be emphasized in the project.

The research project focuses on strategy detection of agents in general. However, the RoboCup is an ideal test-bed for the methods to be developed. The technology of strategy recognition is of large interest in two research areas: on the one hand, the quality to forecast the actions of physical agents can be increased, which plays an important role in the context of controlling autonomous robots, on the other hand, it can be employed to increase the robustness and security of electronic markets.