Óòøöóðððòò Øøø Ãããôöö Êóóóø Ý Aeaeùööð Aeaeøûóöö Åóùðð×

In this paper we present neural network modules that are capable of controlling the miniature robot Khepera. As an interesting application we decided to tackle the task of playing golf with the Khepera. Several challenging subtasks have to be fulfilled to perform the complex behaviour needed. Our special focus of interest lies on: building a map, localization within the map, hitting the ball and local reactive path planning using neural network approaches. From these topics of work we present the basic principles we have used and some of the results achieved within the experiments with the real Khepera. The realized capabilities showed encouraging results for controlling autonomous robots with neural network modules.