Today we started with the round robin games. At noon we placed against the Hamburg Bit Bots. Though we managed to move the ball towards their goal, and sometimes even quite close, stability issues caused our robots to miss all chances to score. At the end, the game ended with a 1:1 tie (both goals scored by Hamburg).
The second game went significantly better, though there is still room for improvement. We played against the WF Wolves and the final score was 5:0.
At https://www.facebook.com/fumanoids you can see two short videos of Emmy scoring a goal (once in a test match against the Bit Bots, once against the WF Wolves in the round robin game).
This evening we plan another test game against the Hamburg Bit-Bots, and tomorrow at 11am we will have our final round robin game against the Bold Hearts.
Today was the second setup day. We started the day with an extensive check of the robots, improving various aspects and preparing for the second test match against the Hamburg Bit Bots. This game went much better (for both sides) than yesterday, with a final score of 1:0 for us (pictures below). Obviously this still is far from perfect, but these games allow us to spot the most critical points we need to work on and we hope that tomorrow in the official games we will score more goals. Maybe we’ll have a third test game this night.
In the afternoon we had the referee training, and afterwards our robots were inspected for rule compliance. Steve unfortunately did not pass the inspection but all other robots are allowed to play.
We now also know the schedule. As there are only 4 teams participating after all, everybody will play everybody in the first round, with the first two games tomorrow and the third game Friday morning.
Thursday 12pm: FUmanoids vs Hamburg Bit-Bots
Thursday 4pm: FUmanoids vs WF Wolves
Friday 11am: FUmanoids vs Bold Hearts
On Friday afternoon will be the semi finals (at 4pm and 5.30pm) and on Saturday we will have the final at 2.30pm and the game for 3rd place at 1.30pm.
We are off to participate in the RoboCup German Open. Tuesday and Wednesday will be setup/preparation days, and Thursday to Saturday we will have the actual competition. If you are in the area, swing by to watch a game (free entry).
Updates will be delivered via Facebook (no account needed to view our posts) and hopefully, time permitting, via this blog.
If you followed us on Facebook, you already saw how our sixth robot – Emmy – was coming together. It was a long way for Emmy to start walking. Initial plans for here were made over a year ago, when we manufactured enough mechanical parts for six robots. In January, Emmy’s legs and torso were assembled, however we couldn’t finish as we did not have enough working servo motors for the head and feet. The motors we had left didn’t have working boards, something which we could not easily repair ourselves. Thankfully the Darmstadt Dribblers helped us out – as they replaced all their servo boards they kindly provided us replacement boards for our servos. This finally allowed us to finish Emmy (and still have a few spare servos).
Emmy's legs and torso were assembled in January/February. At this time we did not have enough working servo motors to finish.
Testing computer vision, modeling and behavior on a humanoid robot can be difficult. At some point you can’t rely on the simulator anymore and have to test on a real robot. And suddenly everything becomes more complicated and the results less reliable. Swearwords can be heard when the robot is not working perfectly (lose screws, mis-calibrated servos, …) and time has to be spent doing basic repairs and configuration. Even if the robot is in perfect condition, the change from simulator to actual walking robot is a big difference due to a lot of noise in the image and sensor data caused by the robot walking.
So for testing, we already showed you a three-wheel platform for our robots a while ago. After we had to recycle the servos in it for repairs, it got a bit forgotten until a couple of months ago when we updated the design to now be a standalone robot. Luckily we saved the prototype of the torso which is now mounted on a new base platform. We also updated the omni-wheels (again borrowed from one of the FU-Fighter small size robots) to a smaller version to prevent judder while driving. Adding the head, we now have Steve, whose camera is precisely the same height as our bi-pedal robots and which can run the same software. As it only has 5 servos, the battery lasts much longer which is an added benefit when testing. And as Steve can not fall down and has not many moving parts, it is (almost) always in perfect working condition.
Overall, this makes for a very nice test robot. Check out the pictures below.
Steve, our non-bipedal humanoid test robot.
Steve additionally has a kind of hat, which the groundtruth system in our lab (two webcams mounted on the ceiling) use to track his position on the field. This allows to analyse the accuracy of e.g the self localization or ball model.
The motors are Robotis RX28 models. We used three servos, which provides sufficient power, however for faster movement it should better be four servos.
The wheel was kindly provided by a FU-Fighter small size robot. The coupling was specifically built in the workshop for Steve.
Steve from the back. The torso prototype did not have the final design, so we could not mount the full button panel at the back. But power on and off is available at least.
Steve from the back again. You may have noticed the missing arms, however to save weight (and servos) we didn't add the arms as Steve has no use for them (they are not required for driving and he does not fall down)
Steve and Grace. The upper body is the same height above ground, which makes it easier when working with the camera projection.
After having coded past midnight on Friday, the teams arrived back this Saturday morning to prepare for a day of coding, talking and tests. We did our first friendly game this year, against the Hamburg Bit-Bots. Some issues on both sides caused the number of shot goals to be quite low (final score was 1:1). , however that’s what these games are for. Overall we are quite optimistic about the state of our robots and software and are going to work on some of the issues. The Bit Bots played against Bold Hearts as well. Hopefully we’ll have more games tomorrow.
This weekend, we are hosting the latest of the RoboCup Berlin Open Workshops (“RoBOW”). While the SPL-Teams meet in Dortmund, several Humanoid teams have come to Berlin to discuss the latest developments and test their robots in some friendly games. What a wonderful opportunity to test the new big field and prepare for the upcoming competitions in April.
After numerous complaints (not just from Ada and Grace), it became obvious that the robots needed new shoes. The old ones were quite clunky and also heavy, so a new much lighter model was designed and added to all robots.
In the spirit of sharing, our team has made available all developed electronic components used in the FUmanoid robots. This includes schematics and board layouts, as well as the required firmwares.
The first component released is our power board. It includes a microcontroller to monitor voltage, report the system state and cut off power if necessary.
Next is our sensor and actuactor board called Erolf. It provides three serial buses to the servo motors (in our case RS485 Dynamixel models), allowing for high-speed parallel setting and getting of the actuator position. It also includes an IMU with a dedicated microcontroller running an extended kalman filter. The board also includes safety functionality to cut off the power to the servos as soon as an abnormal state, e.g. a short circuit in the cables, is detected.
Erolf – an actuator control and sensor board
Finally, the third component released is an overvoltage protection module for our Dynamixel servos. Though quite small, it serves an important functionality in that it short-circuits the power to ground (detectable by our power board) as soon as a short circuit is detected between the data wires and VCC. This usually would cause many of the connected servos to become damaged (more specifically, the ICs connected to the serial bus). Together with the above two boards such damage is now virtually eliminated in our robots.
February 20th, 2014 | Category: Uncategorized | Comments are closed