CS Freiburg: Doing the Right Thing in a Group
The success of CS Freiburg at RoboCup 2000 can be attributed to an effective cooperation between players based on sophisticated soccer skills and a robust and accurate self-localization method. In this paper, we present our multiagent coordination approac
- PDF / 279,126 Bytes
- 12 Pages / 451 x 677.3 pts Page_size
- 2 Downloads / 188 Views
stract. The success of CS Freiburg at RoboCup 2000 can be attributed to an effective cooperation between players based on sophisticated soccer skills and a robust and accurate self-localization method. In this paper, we present our multiagent coordination approach for both, action and perception, and our rich set of basic skills which allow to respond to a large range of situations in an appropriate way. Furthermore our action selection method based on an extension to behavior networks is described. Results including statistics from CS Freiburg final games at RoboCup 2000 are presented.
1 Introduction After winning RoboCup in 1998 and coming in third in 1999, CS Freiburg won the competition in the F2000 league at RoboCup 2000 again. One of the reasons for this success is most probably the accurate and reliable self-localization method based on laser range finders [10]. However, while this was basically enough to win the competition in 1998, it was necessary to work on a number of different problem areas in order to stay competitive. Our main research aim this year was to develop and test new techniques in the areas of action selection and multi-robot cooperation. In order to do so we also had to redesign the action repertoire and to rethink what kind of problems could be solved by a group of robots in which way. In addition to these issues we further enhanced our perception technology. However, most of the work in this area was already done last year [13] and will be described only briefly. Figure 1 depicts the software architecture of our players. The perception technology as described in Section 3 is the basis of our team. In the area of cooperative sensor interpretation we were able to come up with interesting and significant results. As described in Section 3.3, the group estimation of the ball position is done using a combination of Markov localization and Kalman filtering. For RoboCup 2000 we completely redesigned our strategy component. We employ a variation of the dynamic role assignment as used in the ART team [5] and a variation of the SPAR [14] technique for determining optimal positions on the field for each player. One of the objectives in the development this year was to enhance the basic skills of the soccer agents. In particular, ? This work has been partially supported by Deutsche Forschungsgemeinschaft (DFG), by
Medien- und Filmgesellschaft Baden-W¨urttemberg mbH (MFG), and by SICK AG, who donated a set of laser range finders and supported us in constructing a new kicker. P. Stone, T. Balch, and G. Kraetzschmar (Eds.): RoboCup 2000, LNAI 2019, pp. 52-63, 2001. c Springer-Verlag Berlin Heidelberg 2001
CS Freiburg: Doing the Right Thing in a Group
communication
sensors
strategy
action selection
perception
basic skills
53
actuators
Fig. 1. Player architecture
we added a dribbling skill and a ball-shooting skill. For both of these new skills, it was necessary to modify the hardware and to account for these modifications by incorporating new software. In order to choose the right action in
Data Loading...
