Strategic Control of Mobile Robots

We have developed a method of representing robotic soccer game. This representation is used for controlling robots playing soccer. Our approach to robot soccer is to view it as a local interaction game, which reduces the need for centralized control. In this paper we describe both hardware and software implementation that is used by robots. Virtual grid is described. The virtual grid is much more efficient than representation of game using natural coordinates of robots. This representation can be efficiently used for a description of game strategies, which can be directly used for a faster control of robots. We also show how a can be learned and illustrate our approach using the latent semantic analysis. Strategy learning is important both for developing successful players and also for discovering strategies of the opponent team.