We have improved on the original implementation. The possible update rate has been improved from the original 80 Hz to a theoretically possible 9.5 kHz. Our transfer rate through the bus is limited and thus we filter the output from the load cells and only transfer with a speed of 1 kHz.
We have put the Sensors on the top side of the feet of our robots.
We have implemented the sensors to help us gather more data and use this data to improve our walking and our animations by making them more stable.
We have put the sensors on our robots on the RoHOW. The sensors are able to measure 40 KG each without breaking. Of course we took that as a challenge and used two thumbs to create as much pressure as possible. There was not enough space near the sensors to use more than just two thumbs.
In the future we plan to scientifically compare how our measured center of pressure compares to an industrial 6 axis force torque sensor and publish our results. In the near future we also plan to release our improved version of the ForceFoot as opensource hardware.
In the last few days, interested teenagers got to try out, what it is like to study computer science at our university. And like the last years, we offered one of the three projects the pupils could choose from. During this week, they did not just get a look at our daily lifes, but instead got to experience hands on, what it takes to make a robot play soccer.
Our new friend Cozmo, a robot produced by Anki, joined us for this years project, and thanks to its intuitive programming environment made it rather easy for the students to solve various problems and to programm and play games with or against Cozmo. Step by step they helped Cozmo to reach his dream goal to become a soccer star player.
In contrast to the last years, this time we tried a different learning concept, developed by a master student at our university. In various tasks the 13 children first learned how to work with the robot and his graphical programming environment, before later going on to use a “real” programming language. The main goal was to make Cozmo shoot a goal on friday.
Even though this was a challenging task, the students quickly recognized how to use Cozmos sensors to find the ball in the camera image, then calculate the position in the real world, to then figure out where he is supposed to move.
Besides working on the project, the pupils were also keen to see our campus, so we showed them around in the labs of TAMS. They got to control a robotic arm with a HoloLense and watch as Trixie searched for interestingly colored objects.
Even though our last article was a while ago, a lot has happened. After the group phase we were eliminated from the regular competition of the world cup in a match against Rhoban from France.
We were more successful in the Drop-In Challenge: because of our stable goal keeper we managed to finish in the 4. place.
Afterwards we supported our collaboration partners in the team WF Wolves from Wolfenbüttel, who played in the TeenSize League. They finished the group phase in the third place and thus got directly to the semi final. There we cooperatively lost against the team Ichiro from Indonesia. The match for the third place against the Nubots from Australia ended in a penalty shootout which we unfortunately lost. In the Drop-In Challenge in the TeenSize League we were more successful and managed to finish in third place.
In the KidSize League Rhoban has managed to become World Cup champion for the third time. In the TeenSize League Ichiro became the World Cup champion for the ninth time.
After the competition many members of our team spent a few more days in Canada before we went back to Hamburg. In the near future we plan to extend our cooperation with the Wolves. In September they will visit us and we will play test games and work together on scientific papers.
The last two days were accompanied with intense preparations: At 8:00 AM we arrived at the exhibition on saturday, in order to prepare our tables, repair our robots, which were slightly damaged during the flight, and, as we mentioned in the previous article, to record some images for our vision.
Unfortunately, when we arrived, we had to face an unpleaseant surprise: The registration was not prepared yet, so we got to stand in line, and wait. And wait. Some time later the organizers decided to let us in without the registration (and our badges) for now. The registration was postponed to some time later.
This inconvenience was promptly followed by another minor problem: Apparently, the organizers had a slightly different idea on what the term “Setup Days” is supposed to mean. Usually, the setup days are used to prepare the robots and adjust them to the different circumstances at the venue. This year, however, they were also used to set up the hall: Apart from the junior league (for schools/pupils), no league was completelyprepared yet. In our league, there was nothing except for the wooden foundation for the fields. Everything else was scattered around. There were not even chairs yet! The organizers didnot seem to be able to provide enough manpower to fix this soon, therefore we decided to DIY our fields (as known from this swedish furniture store). Except there was no manual this time. Likewise to products bought from said furniture store, there also were parts missing. Nonetheless, we made up a humanoid-league setup team and took matters in our own hands. This meant we had to lay the grass field, place the markings and perform several other tasks on our own. Especially team Rhoban pulled out all the stops. It still took us until the end of the second setup day to finish all fields. The TC (technical committee) and EC (executive committee) usually are very busy during setup, even without these circumstances, so it came to no surprise when many decisions and announcments about the competition were delayed while coordinating the setup.
One event during the preparation time before the world cup seems to take its revenge now: Pressure from the outside forced the whole local committee to be exchanged, leaving almost every position with someone inexperienced.
The organizers are mostly experiencedin the junior leagues only and the required materials and man hours were vastly misjudged. Especially severe, as the participation fee for the world cup is not minor and thus setup infrastructure and prepared fieldswould be expected.
Daniel Polani (President of the RoboCup Federation) at the Opening Ceremony
When the world championship was officially started with the opening ceremony on sunday, the president of the RoboCup federation found proper words when he thanked the attending teams for their effort in setting up the required materials and fields.
We are confident that this kind of chaos will not occur again at future RoboCup competitions, because the RoboCup federation will hopefully find ways and means to to support future organizers and demand better terms for the teams.
Marc checks the walking
Regardless of all that, our team, of course, also worked on our robots and software. Seeing how our robots walked was great, because of how stable it is compared to last year. They are now able to walk all across the field without any trouble. On the first setup day, Marc was the first to put a robot onto the field.Accompanied by astonishment from the whole team, the robot just strolled across the field. He was not even bothered by any twists and turns or a side step. What a difference to last year! This strongly contributed to the motivation of our team. A lot of work over the last year went into our vision and ball detection, which also has paid off. We were also able to demonstrate our ImageTagger to other teams and make it attractive to them. The paper that was written about the ImageTagger by Niklas, Marc, and Norman Hendrich was also accepted for the RoboCup Symposium (a short conference ensuing the RoboCup). Other than that, we are working on all kinds of small problems and are looking forward to the rest of the RoboCup.
It’s unbelievable but almost a year has passed since the last world championship in Japan has ended and the world championship in Montreal, Canada, is just around the corner! This year, the schedule is a little unusual: The set-up days are Saturday and Sunday and the regular tournament doesn’t start until Monday.
Like last year the arrival of the team members happened in chunks: Jessica has already arrived yesterday and used the day to explore the city. Everyone else traveled today: Maike’s flight from Stockholm left at 06:05, Judith, Jasper, Marc, Timon, Niklas and Daniel left Hamburg around 10:10. They re-united in Paris and started the last (and biggest) part of the journey from Paris to Montreal with significant delay, but together. The flight was mostly used to work a little on the software, meet new teams and catch up with some sleep. The airport in Montreal was then firmly in the hands of the RoboCup community: Flights from Paris, Amsterdam and Frankfurt arrived nearly simultaneously and brought almost all European RoboCup teams to Canada at the same time.
We then made the first experience with the famous Canadian kindness during the trip from the airport to our accommodation. People here are incredible accommodating and you don’t even need to ask for help, you’ll frequently get kind offers of help. Then again everything is very coordinated here: The shuttle buses from the airport had an “instructor” who decided how many people could board the bus. This amount was strictly determined by the amount of seats and luggage space available. What a difference compared to Japan where people got payed to push as many people as possible into one train!
View towards Convention Center and river
We have rented a small apartment a little outside the city center this year. This makes the travel to the competition venue a little longer. However, it also means we are all staying together, we can have breakfast whenever we want, we can cook dinner together (if we have the time) – and of course work as long and as much at night as we want! The first arrival at our apartment was surprisingly unpleasant: The code for the front door was wrong and of course none of us had a Canadian SIM card or data roaming available. Luckily the landlord randomly came by after a little while and the short drama had a quick and happy ending. The rest of the evening was used for a short grocery shopping trip and a quick dinner. Now most team members are already asleep, which is no wonder since it’s already 5 AM in Germany.
Tomorrow morning the competition hall will open at 8 AM and of course we want to be among the first to enter. We hope to record some images to train our vision and adapt our walking algorithm to the local turf right away. Much has changed since the last world championship, both when it comes to the software and hardware. The level of motivation in the team is higher than it has been for a long time and thus we’re all looking very much forward to the beginning of RoboCup 2018 tomorrow!
Convention Center, Palais des congrès de Montréal Robocup 2018
The IranOpens are over and everything is back to normal again. Time for us to look back at the last few weeks and draw some conclusions: Many of our modules have been used for the first time and we could gladly see, that most of them worked even better than we expected them to do under real life circumstances.
Our new vision and walking algorithms perfectly work together with older modules and brought us some very successful plays. We found and removed many huge and tiny mistakes and bugs, that existed undetected for a long time. The exchange with the other teams broadened our horizont, gave us the possibility to discuss human-league problems, inspired us to new ideas and let us make new friends. Last but not least, this years IranOpen was the first chance for three of our members to experience working at a RoboCup competition.
Of course, we also realized what doesn’t work that well yet and we are working on fixing these problems as soon as possible. We are mainly focusing on getting our new robots to work, so that we can run four robots at the same time for the world championship, which draws nearer and nearer. We are looking forward to see our robots score some goals.
All in all the competition was a huge success and lets us confidently work towards the world championship.
Our last day started as chaotically as we already anticipated it to be: 10 minutes before our semifinals-game was supposed to start, the exhibition halls doors were still locked. As the referees and the TC didn’t show up on time aswell, luckily we still had enough time to prepare.
Playing against Rhoban was a challenge, as always, but with our vision working perfectly and optimized walking algorithms we did pretty good. A few minutes before the game was over chaos took over once again: Due to unknown reasons the power was cut and therefor all game events stored in the Gamecontroller were lost. The referees and the TC took some time to deliberate about what to do now and then decided to continue the game anyways.
In the end Rhoban beat us, leaving us on the third place out of eight competitors. The first place goes to MRL-HSL.
After the games were over, we had some time to visit a temple with some members of Rhoban and ZJUDancer, and to show them around on the bazaar. The rest of the day was used to attend the very exciting closing ceremony. After the banquette, where we could once again talk and joke with our – sometimes long known, sometimes new – friends, it was already time to say goodbye.
13 hours of flight later, completely exhausted but very very happy, Hamburg had us back. We won’t have much time to rest on our success, though. The world championship is getting closer and today the lab was as crowded as always again.
Today began as early as the days before. Our first game against MRL-HSL was scheduled for 8:30 AM. Until then we still had to fix some hardware components. As we were working on our code until late in the night, one could see our exhaustion in our faces. Nonetheless we played a good game, which, unfortunately ended in a loss, but proved that our hard work during the last days was a success.
Our robots fell very badly during the game, damaging parts of our hardware beyond repair and causing major problems during our second game. Therefor Frankenbot now inherited Minibots arm aswell, making him more and more a patchwork-beast. Between the other teams, Frankenbot made himself a name as “Killer” or “Terminator” and is well known. Because there is one thing we could prove during the last days: We are very good at not getting tipped over.
In our free time after the games we went out to explore Teheran. We bought lots of food, spices and sweets at the bazaar and visited a mosque nearby.
During our third game the hardware problems lingered. By now everything seems to be working again and we hope, our robots won’t get hurt again tomorrow.
After the games we had some time to talk to the other teams. We introduced our ImageTagger to the other teams, Rhoban held an interesting presentation about decisionmaking processes during goalshots and ZJUDancer and MRL-HSL showed a part of their works aswell.
We are now making our last preparations for our semifinal-game against Rhoban. We are curious to see the outcome.
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.
