RoboCup International Competition 2004 - Lisbon, Portugal

[patrickrd]

Hi everyone,

Like last year, I will be posting updates about the international RoboCup competition. This year, the event will take place from June 27 - July 3, in Lisbon, Portugal. I will specifically focus on my team (Cornell University "Big Red") and the Small-size ("F-180") league, as that is what I am directly involved with.

RoboCup is an international non-profit organization that promotes new technologies and advancements in robotics through an annual competition and conference. The long term goal of the organization is to field by the year 2050 a fully automous robotic soccer team that can compete against the human world cup champion soccer team. At the current time, RoboCup contains 5 leagues, which each address specific aspects of playing soccer:

• Simulation league: Simulates soccer matches. Works on high level soccer strategy and individual player strategy and skills.
• Sony AIBO league: Implements soccer skills and strategy into AIBO dogs.
• Small-Size League: Makes fully autonomous miniature robots that play soccer with a golf ball. Most competitive and advanced league in RoboCup.
• Middle-Sized League: Makes larger robots that play on a larger playing field. Performance is limited due to vision restrictions.
• Humanoid League: Makes humanoid (legged) robots. Very research oriented at this time.

As I mentioned, we (Cornell) compete in the small-sized league, which contains the most exciting and high-paced games in RoboCup. We have won internationally four of the past five years, since our first year in 1999. Last year, we won internationally in Italy, and I posted updates of the 2003 competition here on Chief Delphi.

This is the 2004 thread, where I will post updates from Lisbon. In addition, I will be posting some updates about how our team is doing before we leave for Portugal, and some of the new things we are introducing. Be sure to check this thread often and wish us luck!!

[Andy Grady]

I once saw a piece on Scientific American Frontiers on Robocup, and I must say I was extremely impressed. I wish you and your team the best of luck! Will any of this be televised in any manner?

-Andy Grady

[patrickrd]

Originally Posted by Andy Grady

I once saw a piece on Scientific American Frontiers on Robocup, and I must say I was extremely impressed. I wish you and your team the best of luck! Will any of this be televised in any manner?

-Andy Grady

I don't think it will be televised, at least not live. All the games are recorded though, and end up in all sorts of places. I just put this thread in the Televised Robotics forum because I didn't know where else to put it.

[Yan Wang]

Good luck as always.

And I want my jersey.

[patrickrd]

Today is two days before we leave for Portugal. Our team is doing well. Everyone seems to be in good spirits, although we have a lot to do.

Mechanically, we had a major setback about two and a half weeks ago when our only CNC Mill went down. It took a week and a half before it was at all usable. One other team member and myself have been switching off, running the machine up to 16 hours a day. In fact, the machine is running next to me right now, as I write this update. The good news is that we already had 5 robots before the machine went down. Right now we are machining new parts for the robots. A new revision of most parts will be going on the robots this evening. There have been few mechanical problems with our robots this year.

Electrically, we have also had a number of setbacks. Our electrical system is very complex this year (see below). There have been many problems with wireless, motor/servo control, and board problems. Most of these problems our EE's have resolved, but it is still a very difficult system to keep up and running.

AI/Strategy/Control: We are quite behind. However, the important stuff (controlling the ball, moving quickly around field, shooting) is improving quickly. Our strategy has been implemented through simulators over the year, so that mainly done. However, we will want to tweak it to utilize new skills that our robots have, and based on the performance and capabilities of other teams.

Here is a list of new things that our team is adding to our robots this year:

- On board computer: Each robot has a 400 MHz processor on it, running windows CE. The robot uses compactflash for memory.

- Intelligence moved onto each robot: For the first time, Cornell is moving all the intelligence onto the robot. Each robot will independently have the capability to decide what to do. In the past, a computer off the playing field has made the decisions and sent commands to the robots. This year, the only computer used off the playing field interprets a vision camera, which looks at the playing field and determines where all the robots are and where the ball is. This information is sent to the robots over our 802-11 wireless system.

- Bluetooth wireless communication: Although each robot has the capability to independently decide what to do, the robots are also capable of communicating directly with any other robot(s) via a bluetooth local area wireless network. This is analogous to humans communicating with each other while playing soccer ("time", "pass here", etc...). This network is also used to send software updates to robots without the need to take them off the field. Of course, updating software is not allowed except during a timeout.

- Dribbler modification: We removed the side dribblers, which allowed us in the past to move laterally (sideways) with the ball at high speeds. This was removed due to rules changes. Instead, we designed a "notched" dribbler, which tends to keep the ball in the center of the robot. While it is in the notch we can rotate, go forward, backwards. However, moving laterally is not as effective.

- Proximity sensing: Each robot has two infrared sensors. The two beams scan the front of the robot. If we have possession of the ball in the notch, both beams are broken. If we have possession to the right of the notch, one beam is broken. If to the left, the other beam is broken. This allows us to know where the ball is with respect to the front of the robot. As a result, we can quickly move the robot to keep the ball in the center.

- Local robot camera. Each robot has its own camera, facing the "front" of the robot. The purpose of the camera is identical to proximity sensing, however the camera gives us more data (exact location of the ball). It can locate the ball anywhere in near the front of the robot. This allows us to control the ball with great precision. However, this technology was just implemented this past week and it is not clear how much we will be able to utilize it given we are leaving in two days.

- Chip kick: Our robots have the ability to kick the ball in the air. However, we have not yet implemented into our strategy. Our kicks are also more powerful.

- Brushless drive motors: Our robots can accelerate and move even faster than last year. Although we have not measured, the acceleration should be close to 1G (8 or 9 m/s/s) and max velocity can exceed 3 m/s.

- Goalie: We have developed a separate robot this year, designed specifically to be a goalie. This robot can intake the ball, bring it to the top of the robot via a belt system, and launch it to midfield via a solenoid-powered arm. This technology is also very new, so it is not clear whether we will be able to implement it at the competition.

- Strategy changes, to be announced.

Okay I've got to go, so that's my update. I may post one more before we leave for Portugal on Saturday.

[Andrew Rudolph]

sounds awesome, good luck to you guys in portugal!

[patrickrd]

Hey everyone,

Sorry for the late update, I have been unable to get internet access until now. But even if I had, I wouldn't have had time to update. Anyway, here's the latest. I hope to be able to update every evening.

So on the 27th at about 8AM we arrived in Lisbon. Got to the venue quickly and with no major problems or delays. The first day of setup went by very quickly. We got everything unpacked, and attempted to run the system. Our first major problem encountered was vision. The field conditions here are very suboptimal. The lights are yellow (and do not span the whole color spectrum) and our vision system needed major work. Meanwhile, we assembled updated robot parts, and assembled our spare robots. Electrically and mechanically, we also had frequent failures.

On the 28th (yesterday) we attempted to get our whole system up and running. No such luck. Mechanical and electrical failures kept CS from getting a full team of robots. Further, vision problems and wireless problems persisted. AI and strategy also needed a lot of work.

Today, we played our first two games in our round robin of 6 teams. Our first game, we failed to get our robots running properly. But so did our opponents, Plasma-Z from Thailand. After a lot of uncontrolled chaos, the game ended at 0-0. We were VERY fortunate not to lose the game in the last seconds. Plasma-Z had the ball and was slowly moving towards our goal along a clear path. About 1 or 2 feet before reaching the goal, time expired and the game ended in a tie. Final score, 0-0 tie.

Our second game was not much better. It was against Lucky Star from Singapore, probably the best team in our round robin group. Wireless problems kept two robots from working, and the other two worked only intermittantly. At times our defense worked well, but for the most part we had little control of the robots. Final score, 10-0 Lucky Star.

So our current situation is not good. However, we believe we can win the remaining three games in our round robin round (two tomorrow, one Thursday). This would likely either clinch the second spot to advance, or tie for the second. If it is a tie, the team that has had the most goals will advance. Thus, we hope to win the three remaining games by wide margins. All our problems are solvable, but it is not clear whether we have the time to do so. If we can win the remaining three, we will be in good shape, having had two extra days to get our system up.

We're in the process of greatly simplifying our system. We are removing problematic components that are not necessary, and making an effort to field strong robots that are simpler and robust. We do not need all our accessories that we have worked on all year in the next few games.

Well that's about all I've got to say. I'm optimistic that we can come back and be competitive.

[Andrew Rudolph]

WOW sounds like you guys are having a tough time, i know we all have faith that you guys can pull it together and get that #2 seed that your looking for. Goodluck and i cant wait till the next installment!

[lynca]

Good Luck in you matches! The advancements you have made are gigantic.

I have a few questions about your modifications. Are your onboard cameras doing all the work without an outside camera and does the 400 MHz handle the type of processing you require? Also, is the chip kick attribute an important feature to your new system or simply an accessory?

-thanks

Best of Luck to Cornell!

[patrickrd]

Today we did not get the two wins we needed. However, we are not out of the race either. Our first game we tied 0-0. Niether team could get things working well enough to score goals. Our second game was much better, but still pathetic by our standards. We fielded five robots for a portion of the game for the first time, and our robots were much more controlled (but very slow). We won 3-0. We're fixing problems slowly, but I think it it will be a long shot to make it to the finals. In fact, there is a good chance we do not advance past the first round.

In order to advance, we need to be among the first two teams of the six in our round robin group. Lucky Star is clearly the most functional, and Plasma-Z is also relatively functional. To advance, we must win our game tomorrow, and the Canadians must beat Plasma-Z. That would give us an outright 2nd place and we would advance. If Plasma-Z ties the Canadians, then we would be tied with Plasma-Z for second place at 3-1-2. The team with more total goals would advance. Right now we are three goals behind their goal total, so we would need a good game tomorrow. Unfortantely, we cannot control the outcome of the Plasma-Z game. Ironically, Plasma-Z has robots with nearly identical mechanical design as our robots last year (which is great, sharing information is encouraged in RoboCup).

Well, we are having a lot of electrical problems. Our robots (with 400 mhz computers onboard each robot running Windows CE) are very powerful and have a lot of features. They break down a lot with problems very difficult to diagnose. Mechanically, we solved a problem with kicking failure, and our system is quite robust right now requiring rare maintenance. In our spare time, we completed our goalie robot, which works great. But we may or may not introduce it to the game tomorrow, depending on the importance of the game and what other things need to be done. Major control problems, vision problems, and AI problems seem to still exist.

Today Portugal beat Holland 3-1 in the semifinals of the EuroCup soccer tournament held here in Portugal. There are people running all around singing, horns honking, whistles blowing. It's an amazing time to be here during that tournament as well.

[patrickrd]

Quick update: Plasma-Z lost to the Canucks this morning, thus we are now in control of our own fate. If we win against the Canucks six hours from now, we advance. Otherwise, we're done.