The first match day is coming to an end. Our enemies being three of the four bestplaced teams in last years world championship, our competition couldn’t be much tighter: We’ve got to face MRL-HSL, Rhoban and ZJUDancer. Especially Rhoban demonstrated clever teamplay and clean goalshots with their three robots and beat all three enemies. The other two teams played very good aswell, leading to many exciting moments. Even though we lost all three of our games today, we made great advances aswell: For the first time since our big hardware modifications we are able to walk long distances again. Our ball recognition works perfectly and huge parts of our behaviour are already functioning normally.
Playing against some of the best teams in our league was a good oportunity for us to learn. The knowledge exchange between the teams and the games provide great insights into new tactics and different gamemoves, many of which we are looking forward to try out. The huge amount of games also provides lots of data for our image tagger under realistic conditions.
The hardware wasn’t working in our favour today, and so Frankenbot happened to lose a leg – just a few minutes before our last game was about to start. Luckily the damage could be repaired quite easily. Unfortunately we needed one spare part, that we couldn’t replace. So we had to remove it from the other Minibot, which means, that we can only use one of them for the rest of the competition. At least we can focus our work on that one now.
We are now working on fixing the behaviour completely and optimizing some parts, in order to get better results in our next games.
With loads of luggage and close to no sleep we finally arrived in Teheran. We didn’t get much time to rest, because just two hours after the arrival at our hotel we had to depart to the venue and begin with the setup.
Everything seems to be a bit more organized in comparison to our experiences in the last years. Just a few hours later we actually had power and internet access, and we could begin our preparations. Lucky for us, as there were lots of things to be done: Besides having to prepare our two Minibots, this year for the first time we bring two new participants to the IranOpen. While we expect the two new robots, produced by the WF Wolves, to improve our performance, the new platform also brings some new difficulties with it.
By now the damages caused by the flight are fixed and our software is on a good way to be up and running by tomorrow morning. Our robots have been measured and this year we also got a (very brief) referee training. We will now use the rest of the day to keep working on our software, so that we will be ready for our first game tomorrow at 8:30 AM. We can’t wait to see what the following days will bring.
On tuesday, the 20th of February, the annual University Day took place at the UHH. While we couldn’t participate last year, due to the foyers being closed down, this year we could demonstrate our robots and inspire young students to study a computer sience related topic.
Interested students had the chance to steer a robot themselves. Even though Tamara (one of our Darwin-OP robots) usually works autonomously and makes her own decisions, this time we decided to put a leash on, or rather a controller.
The new semester has begun and with it our preparations for next year’s contests. Even though our reportings on this site came to a full stop after we dropped out of the world championship, we kept conversing and interchanging with other teams. Our presentation about the new ROS-architecture during the symposium also was a great success. After a short recreation-phase in the summer, we went back to our normal shedule by September. Our first project in the new semester was a cooperation with the “Gymnasium Walddörfer”, a secondary school in Volksdorf. Together, we hosted a school project week for students in the IT-profile and let them have a peek into the field of robotics. In October we prepared a part of the trial studies, where prospective students could try out studying in computer science during their holidays. During the same week we presented two of our projects at the “Informatik-Expo”: The new ROS-architecture in our league and the ImageTagger, a tool for labeling images.
We also participated in the “lange Nacht des Wissens” again, this year at a panel discussion of the legal studies faculty.
In December we will visit Bordeaux for a weekend to take part in the 5th international Humanoid Soccer School, hosted by team RHOBAN, the current world champion in our league. We’re happily looking forward to an lively exchange.
Today the round of sixteen started. We played our match against the NUbots from Australia. As our walking algorithm was hard to configure to artificial turf, we tried to include the walking code from Rhoban from France, which is open-source, into our software. We managed to do so just two hours before the game started and so we did not have much time to test it. But during the game the improvement was clearly visible. Probably we are going to develop it further and adapt it to the specifics of our robots, as it is easier to maintain than our old walking algorithm.
The game ended without goals, so we had to do a penalty shootout, which the NUbots won with 2:0. So they entered the quarter finals. For us, this is the end of the competition, but now we have time to talk even more to other teams to exchange experiences.
In the drop-in games we are ranked 4th place. As our robots didn’t score any goals by themselves, this result is mostly due to our (great!) teammate-teams, and the factor that our robots were fit to play and so were on the field most of the time.
After being briefly introduced and tried out in RoboCup 2016, this year officially included the Drop-In Games. Four teams bring one robot each to a joint team. This encourages teamplay between teams as well as cooperation. These games are especially challenging for the referees as they have to look out for eight robot handlers (one for each robot). For the teams it is interesting to see which team has which strength. We believe that you can get a good overview of the league through these games. The results are scored in a separate scoring system. Each robot gets points for being in play while a goal is scored and more points for scoring a goal themselves.
The regular games started with a preliminary that determines the composition of the round of sixteen.
Yesterday we played SEU-UniRobot from China (loosing 0:1 to them) and ITAndroids from Brazil (ending in a 0:0). Today we played MRL-HSL from Iran loosing with a 0:1. We now stand at 3rd place in our group and we will play the NUBots from Australia tomorrow.
Today we were able to use our “new” robot for the first time. In actuality it is a Minibot with newly manufactured parts for the legs and arms and an old torso. The missing metal parts for the torso which we replaced with old parts were meant to arrive at our lab about a week before we took our flights to Japan but due to difficulties at the post office we never received them and they were sent back. After assembling him here at the venue we got him through robot inspection and he even received a name: We call him Frankenbot.
Furthermore we used our time to exchange with other teams about different robot platforms, since we are looking for a new one ourselves. New materials such as carbon fiber are especially interesting to us.
It was very relaxing to get some sleep over night in a comfy bed and after a short breakfast (cereals with milk), we started our way to the competition halls.
Travelling by train gets easier each time, the biggest difficulty is to find the correct price for the station you want to travel to. Then you just buy a ticket with the corresponding price and use it to enter and leave the train system. For the next day it will be even easier as the organisers will be handing out train tickets to RoboCup participants so that we don’t have to queue at the ticket machines anymore.
Arriving at the venue we quickly saw that the organizers prepared a lot to avoid mistakes of some previous RoboCups. The registration area was clearly laid-out and it took us only a few minutes to get our team badges and train vouchers. We found the competition fields to be fully prepared, too, and quickly started to set up our team tables.
Next on our schedule:
Calibration of the camera to adapt to the light in the hall
Making sure the robots survived the travel fine and are able to start the system
It seems like yesterday that RoboCup 2016 was in Germany, the world championship on home soil – and suddenly the next world championship is just around the corner! In 2017 it takes place in Nagoya, Japan. While Maike was the first to arrive in Nagoya already on Saturday, almost all other team members finally arrived today as well. Tomorrow, just in time for the start of the two set-up days, Jasper will be the last of our team to join us and will thereby reunite our team again.
After a total travel time of more than 12 hours, first to Helsinki and then to Nagoya (for Daniel and Niklas with an additional stopover in Tokyo), everyone is totally tired. Nevertheless, we still had enough energy to explore the surroundings of our hostel near the central station in Nagoya at least a little bit. As a member of the organizing committee, Maike was already able to inspect the hall for the very first time. All in all, it looks very promising, even if our robots will likely have some problems dealing with some gaps in the artificial turf. However, we will only be able to investigate this more tomorrow, when we are all permitted to enter the hall with our whole equipment.
After a traditional Japanese dinner, the suitcases are now packed with the equipment for tomorrow and we are looking forward to starting a new RoboCup adventure at 7:30 Japanese time (which is 0:30 German time)!
Since the beginning of 2017 we are heavily working on our new software architecture. We decided to move away from our old self-written system. Starting with the competitions this year, the RoboCup GermanOpen and IranOpen, our robots are using ROS (see ros.org). This gives us lots of benefits but also there is still a lot of work to do.
Most of our code is rewritten or integrated into ROS by now. Since we can use the same programming languages as before (Python and C++), we were able to keep parts of our old codebase. We are going this step together with other teams from the humanoid kid-size league who also are using ROS and the proposed architecture. This enables exchangeability of the ROS nodes. On this way, we were already capable of testing the other team’s vision code with minimal effort on our system.
Other benefits are the debugging tools and other libraries provided by ROS.
We are heavily encouraging other teams also to use our proposed architecture or getting in touch with us to discuss further improvements.
The new architecture was developed as a master’s thesis (see Publications) and will be presented at the RoboCup Symposium 2017 in Nagoya, Japan. We hope that an open architecture can help other teams to focus on improving their codebase and ease the access to the league for new teams.
Today is already the last day of the German Open. Here a short overview of what we did the last days.
We made sure the software was running correctly on all robots so that they are able to play. We worked on Minibot’s walking, and transferred existing motion sequences of Minibot standing up to the new system. Also we put plates under Minibot’s feet, to expand the foot area.
The vision continued working (here a small correction: our vision was not made faster by simply removing features that we don’t need but through algorithm optimisation). We can now change or adapt parameters over wlan while the robot is running
(Black circles are balls that are disregarded, since they are over the horizon, the red line.)
Parts of our working group were busy debugging and revising the GameController. We also worked on a tool to visualise what the robot thinks or knows about the field, like its localisation and where it assumes the ball to be.
Having just one Minibot-robot posed a significant problem in the last days . Since software engineering is focused on the future and thus on software for the Minibot-platform, we need Minibot for lots of testing. Since we worked on vision, behavior, walking and standing-up, several people needed to wait for their turn.
We had planned to already have a second Minibot by now, but due to delays in production sequence we don’t have it yet. In the future it will be essential to have a second (better even a third) Minibot to not delay working on code or testing.
The article about the games from yesterday will hopefully follow shortly after this one.