In an exciting final match in Montreal, Canada, the six-time winner and defending champion took second place with a 1–0 win over its arch rival, Leipzig University of Applied Sciences (HTWK). Both teams had good scoring chances: the Nao team HTWK from Leipzig with its flawless dribbling, and the B-Human team from the University of Bremen and the German Research Center for Artificial Intelligence (DFKI) with its strong and precise shots. “Thanks to the coordinated teamwork of our defenders, we were able to fend off most attacks,” says Dr. Tim Laue, who has been in charge of the team for several years. On the other side, the Leipzig goalkeeper prevented a goal from being scored by B-Human several times with brilliant moves. Leipzig only managed to score on one occasion.
About the RoboCup World Championship
The RoboCup World Championship takes place every year in a different country. B-Human competes in the Standard Platform League of the RoboCup: in this league, teams of five Nao robots from Softbank Robotics each play against each other on a small artificial-turf field with the same hardware, but different software.
New Rules with Free Kicks and Goal Kicks
There were a few innovations to the rules of the game this year as well. Most changes had already been implemented at the RoboCup German Open last April, including the introduction of free kicks and goal kicks, which significantly increased the demands on playing as a team. At the same time, the usable bandwidth of Wi-Fi communication between the robots was reduced by almost two-thirds, which meant that team communication now had to be much more concise. Fouls were also penalized more severely, as the duration of time penalties increased sharply with each additional instance.
Additional Events Such as Penalty Shootout and Mixed Team Competition
Just like in Japan last year, there were additional events in Canada, such as the penalty shootout, as part of the Technical Challenge and the Mixed Team Competition. Team B-Human won both tournaments, having taken part in the Mixed Team Competition together with the rUNSWift team from Sydney. Each team provided three players for the joint team called “B-Swift.”
B-Human Uses Deep-Learning Methods for the First Time at a World Championship
Team B-Human currently consists of more than 25 students. Of these, 15 traveled to Canada with us. They were accompanied by the scientists Dr. Thomas Röfer from the DFKI Cyber-Physical Systems research unit, headed by Professor Rolf Drechsler, and Dr. Tim Laue from the multisensory interactive systems working group at the University of Bremen. In terms of content, this year’s team has revisited the topic of deep learning: the training and use of deep neural networks that are used to, among other things, recognize the ball. These new methods were already put to use at the RoboCup German Open in the spring. B-Human was able to record a lot of data and gain important insights to improve various aspects of the system so that the Naos always stay near the ball. The main sponsor of the B-Human team is Bremen-based company CONTACT Software, the leading provider of solutions for the product process and digital transformation.
Additional Information:
Official website of the RoboCup 2018: http://www.robocup2018.com
Official website of the Standard Platform League: http://spl.robocup.org
Website of B-Human: http://www.b-human.de
B-Human on Twitter: https://twitter.com/#!/pennybhuman
B-Human on Facebook: http://www.facebook.com/teambhuman
www.uni-bremen.de
Contact:
Corporate Communications Bremen
German Research Center for Artificial Intelligence (DFKI)
Tel.: +49 421 17845-4180
E-mail: uk-hbprotect me ?!dfkiprotect me ?!.de
