RoboCup 2008 Podcast 1
Leave a reply
Author Archives: Joho
At the Suzhou Expo: Day 1 – Setup
Our day in redux because I am very tired:
We woke this morning at 5:30, got some delicious dumplings and sweet purple sticky rice for breakfast. Took the shuttle to the Venue (20 minutes or so away), our big robot cases just barely fit.
Check-in was a bit of waiting in line, but we got it squared away quickly, and got upstairs to claim our tables. We quickly setup are mobile lab and got developing – Tucker and Henry and Andrew on the Aibo – Elise, Nick working on color tables / picture taking / vision, Jeremy working on Nao systems, and Jack and I on behaviors / Nao vision.
First thing in the morning we got Aldebaran to flash our PSOC camera controllers in hopes of getting a better handle on the camera driver, but to no avail. Instead our robots kernel panicked on boot, and we we unable to use the Naos until very late in the day.
Aibos doing well – orange and yellow are very close, and lots of bars in the images from the frequency of the lights. Otherwise, fields well lit. Internet is very spotty/ non -existent, so we setup our own svn server at the venue.
Naos have a walk that is working, and tomorrow we will get to try a reverted version of our camera parameters, see if we can chase a ball.
Look for a podcast soon.
pSuzhou – the beginnings of something great (RoboCup Day 0)
On the bus to Suzhou from the Shanghai airport, Jack and I started writing what is the fastest behavior ever written – it was created today (Sunday) and will play soccer by Wednesday (knock on wood). Even in 2006, our behavioral code was started at least 2 weeks before competition. It will certainly be an adventure!
Our trip to get here went about as smooth as can be expected. No one got any sleep before our 3:00am meet up time in the Lab, and the 45-min taxi ride in a Lincoln Navigator turned out not be as cramped as we expected, considering the enormous size of our robot cases.
In Portland, Jack and I managed to get away with paying only 80 dollars apiece for our overweight and oversize luggage that could have cost us upto $300 per leg. What a steal! (Mike and Todd had to pay $360, and they’re luggage didn’t even make it to Shanghai on the right flight!) Our 1hr layover in Atlanta turned out to be plently long enough to make it from C55 (where our flight arrived) all the way to the E concourse to catch our 777 to shanghai. Bowdoin RoboCup represented in row 50, while the rest of the plane was pretty empty. The 15hr flight turned out to be just as long as you’d expect a 15 hour flight to be – long enough to take several hour long naps, watch a movie or two, and read some of my book, and still have time to watch the plane inch slowly across the map view in the headrest in front of me.
Once we got to Shanghai, I tried to take money from an ATM, but it said my Bank refused the payment. Instead I found a pretty decent exchange rate right next to the luggage pickup where you pay a flat 59 RMB commission on the exchange – the rate was something like 6.75 RMB/dollar. Once we got our luggage, we were met by some local volunteers who helped us sort out our ATA Carnet, and get through customs. Mainly we had to wait a long time for them to process CMUs Carnet – they never even looked through our luggage, and approved our paperwork quickly. As should be expected with technology, neither of the two cellphones I brought worked when I turned them on. The Bowdoin one refused to acquire any service – maybe it doesn’t work on the 900Mhz band? The other one I managed to borrow from Dave probably needs to get some minutes charged into it – it says Emergency Use Only – at least it gets service!
Exiting from Customs in Terminal 1, we met the rest of the group, who had already been informed by another volunteer that we had another team member waiting in Terminal 2. One of them even helped guide us to the other terminal, which was a relief. In Terminal 2, we met up with Todd and Mike from UT Austin, whose robot’s unfortunately hadn’t made it on the same flight. After grabbing Tucker and Henry, we were again guided to the tour bus, which took about 3 hours or so, and finally dropped off at our hotel. A deluge of volunteers processed all the RoboCuppers on the Bus, and got us meal tickets and room keys. Tomorrow we will start breakfast at 6am, in time to be at the Venue for the 7am opening.
Scrimmage against the German Team, 2-1
Yesterday we scrimmaged the German Team in our lab. The idea is very similar to how we competed at the German Open in April. We sent the German Team a copy of our color table for the lab, and some code that loaded the binary into memory, and Max converted it to the german team’s format.
Our play was comparable to the German Team’s, and we finished the first half leading 1-0 after a lucky shot on goal. The second half ended prematurely after drained batteries and bugs on both sides made play deteriorate, but not until after both teams scored a goal, leaving the final score 2-1. Our major problems came from a bug which keeps the chaser in the grabbing position instead of kicking, resulting in a ball holding penalty, as well as a short episode of ball fright induced by a DEBUG_CHASER switch turned on at half. The German Team’s only issue was an unidentified problem which resulted in the robot ceasing to play soccer and simply swaying back and forth in the middle of the second half.
Unfortunately we were only able to scrimmage 4. v 4. since we our robots are deteriorating quickly. Currently Sam and Pippin are in the shop, and Mike’s leg is about to give out. Even though development on the Aibo is generally straightforward, it is clear that only a few teams (if any!) will be able to muster robots to play again next year if the Aibos are brought back for another year.
Videos from US OPEN are now available
Pool play:
NBites vs. Metrobots (9 – 0)
NBites vs. UT Austin (3 – 2)
NBites vs. CMU (5 -2)
Final:
NBites vs. UT Austin (2 – 1)
Thanks to Nick for putting these together. We also have another video UT Austin vs. CMU waiting to upload once we clear off some hard drive space. All these videos will also soon be hosted on our media page.
First game of the US OPEN: 9-0 against the Metrobots
After an ugly test game last night against UT Austin (formerly the NUbots), we played a really good, clean game against the Metrobots from Brooklyn College this morning. We had a good turnout of visitors watching our match and the adjacent small scale robots.
Footage of the match as well as pictures to come. You can watch last year’s final game of the World Championships against the NUbots here.
We have a match today at 4:30 against University of Texas at Austin. It should be a close one.
Public scrimmage this Sunday May 11th, 2 pm Maine Lounge
.flickr-photo { border: solid 2px #000000; }
.flickr-yourcomment { }
.flickr-frame { text-align: left; padding: 3px; }
.flickr-caption { font-size: 0.8em; margin-top: 0px; }
Just a heads up that if you’re in the area, and you’d like to watch our team scrimmage, we are holding a public scrimmage this Sunday in Maine Lounge in Moulton Union. We will be playing the Aibos on the larger field, 5 on 5, and also have a small demonstration of our new Humanoid robots.
Public scrimmages are a great way for us to get off of our scaled down field in the lab, and onto the regulation sized pitch. Also it will allow us to setup our Nao field for the first time.
Directions here:
View Larger Map
2nd place at German Open 2008
On Friday we finished second in the four-legged-league at the German Open 2008. We played the German Team in the finals, and lost 1-2 in regular time. Last year we finished 3rd, and lost 0-8 to the German Team. Looking over the video of the final, we were still struggling to find and grab the ball, as well as to stay localized on such a large field, leading us to shoot next to the goal sometimes, as well as choose the wrong kick sometimes. Overall, the experience was really good for us – remote competition is definitely a different beast than competing in person, but it does help us pin point many areas for improvement. Take a look at the video of the first and second half.
Getting a chance to play on a larger field in not-so-great lighting conditions is really useful because we can only fit a 72% size field in our lab. Once class get out next week, we’ll move into a bigger space where we can have both the Nao and the Aibo fields out at once. Also, we’re planning to have a public scrimmage the weekend after this during reading period – exact date and time to be announced soon.
With some more work over the next couple months, our second meet-up with the German Team in China is going to be an exciting game.
Northern Bites competes remotely at the German Open
Today kicked off the first day of the German Open 2008, in Hannover. This is the first time we’ve ever tried to compete in an event without actually being there. It is also the first time we’ve ever played on a 5×7.5m field. Although these hurdles might seem trivial, I was ecstatic to see them actually play real soccer in our first test match. To setup our robots, we get images and logs over the internet from the German Team, who have agreed to help us and Cerberus compete remotely.
Despite some technical issues, we were able to produce a decent calibration, and have for the first time, ( to my limited knowledge) played another RoboCup team remotely. In our test game against the german team, both teams remained scoreless after 10 minutes of play, though the German Team came awfully close to scoring on us once or twice. Soon after, we won the first regulation game against Cerberus 7- 0.
Looking at over videos of our performance, it looks like we are having some issues with getting accurate ball distances, as well as some issues with our kick not performing so well occasionally. As always, wireless connectivity is extremely buggy, and it appears our communications were lagging a bit in our game against Cerberus. After some debugging in our lab today, tomorrow should bring better performance in our color table, as well as slightly better role switching.
Read after the break for some technical details about remote competition: Continue reading
Nao setup: first impressions
Out of the box, the Nao didn’t do anything interesting. The boot up time is about 3 minutes, after which the eyes flash, and the processor fan starts. Then nothing. With the robot off, we’ve been playing with some of the joints, and many of them are extremely stiff or choppy. It’s not clear at this point that this robot will actually be able to walk. According to Aldebaran, they will ship us 4 brand new robots in May.
Looking through the SDK documentation, the only way to compile a controller module to actually run on the Nao is going to be on Linux.
All in all, it’s been pretty disappointing. On the Aibo most of this stuff was pretty rock solid, even though OPEN-R was difficult interface to program. Tonight we will try to find out the Nao’s IP address and ssh into it, and see where we go from there.
The only progress we have made is in compiling one of the SDK examples, and having it bind with an instance of the Naoqi running on the mac.
On the upside, the docs show that we will have access to their motion engine, which is ZMP based, and whose parameters we can set.