2 minute read

This year, the RoboCup German Open takes place from March 12 to March 16 in Nuremberg. Not only is that considerably earlier than in previous years, it is also a much longer trip for us. As a reminder: the last German Open was in mid-April 2024 in Kassel. We even blogged about it.

For that reason, and to build up momentum again after our move, we spontaneously decided to hold a team-internal workshop in our lab today.

In keeping with the old team tradition, we met around 10 a.m. in the lab for a minced-pork breakfast and already discussed the goals for the day.

Minced-pork breakfast in keeping with team tradition
Photo: Torben Röhrs

Because of our move to the new premises, we were only able to spend very little time on the robots last semester. Our meetings were mostly spent organizing furniture or moving things in and out. So we used the spontaneous workshop to take a proper inventory for the first time since the 2024 World Championship in Eindhoven. We put the robots through their paces, recorded faults and marked damage. Over the coming weeks until the German Open, these will be fixed systematically, with the goal of bringing all of our robots to the same level.

But our pure computer scientists also used the workshop to test improvements. For the first time, we were able to test our new localization on our field. You read that right – since the move, we have a playing field again! Put simply, localization lets our robots calculate where they currently are on the field. Not an unimportant skill for a soccer player.

Robots on our new playing field
Photo: Torben Röhrs

Besides the work on the field and in the workshop, we also worked diligently on purely digital solutions. A new 3D model was generated so that the robots can be represented better in physics simulations. The model is generated directly from our design tool and then has to be compared against the real robots. And we devoted ourselves to first experiments with the MuJoCo simulation environment – something we already found very interesting at RoHOW 2024 in Hamburg. Building on that, we want to look more into reinforcement learning in the future, in order to possibly replace our walker with a learned model.

Presentation of the new Animation Server
Photo: Torben Röhrs

On the side, we also moved the first robot to our new monorepo. This makes managing the software state on the robots considerably easier, including for non-computer scientists. That was already a problem in Eindhoven, and we hope it will give us fewer inconsistencies in the software state. With the monorepo, we were also able to deploy and test our newly developed Animation Server. It allows us to store and play back predefined motion sequences – like a recording. Movements such as kicking and standing up are realized on our robots through this software.

Overall, the workshop was very successful, and we were able to complete some time-consuming tasks, such as integrating new software packages and taking inventory of the robots. In two weeks the new semester starts, and we are already looking forward to the new students.