Archive for the 'RoboCup' Category
Page 3 of 46
Sorry, our hi-def camera was stolen, so this is the best we’ve got.
That’s the good news. The bad news is that we didn’t play our best game. We led L3M 1-0 at the half, but were off wireless the entire second half which kills our strategy of having the goalie communicate ball information. So we spent much of the half trying to score on ourselves. Our goalie was valiant, but eventually the 4 robots (2 of ours, 2 of theirs) trying to score on her (Zoe) were too much and they scored with less than a minute left.
So our first match tomorrow will be against Penn. They are looking very good and it should be a real challenge.
I expect/hope Wils will provide a more detailed accounting of the day’s activities later when he has some time.
– Aforementioned Detailed accounting –
So we haven’t won a pool since 2009?? That’s really good. We were very upset at our performance though, especially because it seemed out of our control. Our robots started out on the wireless and then completely dropped it. We felt helpless. That said a tie was enough for us to advance and be guaranteed 3 more games so that’s good. In the aftermath of the game we are hoping to develop strategies for if/when we drop wireless again and have also given the router a factory reset which (fingers crossed) should help. We also saw a curious bug that Octavian is looking into, in which the robot decided it would quit our program before each half. Very annoying seeing as we had to call a timeout to reboot them. Hopefully we will track that down. Tonight we have a scrimmage scheduled with our friends the Nao Devils from Dortmund. Hopefully that will look better.
This game started off rough. We had tested our wireless half an hour before the game and were pleased that all our robots were communicating, but five minutes before we had no communication. There was a flurry to fix a bug that we barely caught, 3 minutes before the game but NTU called a timeout and we were able to re-install our fixed code but our communication was still not working. We decided to just go for it with no communication because we can play without it, although not nearly as well. We got manually positioned with kick off as the pink team. Usually the blue team has kick off, but since NTU took a timeout, we got kickoff. Play proceeded for a while, still with no communications between robots… But we scored!! We managed to score on the correct goal without localization or communication. There was much rejoicing. Brian and Octavian were hunting for other wireless networks that could be interfering with ours, but Brian tracked down the problem to a faulty cable that was connected to our own router. As soon as it was unplugged, our robots’ ears lit up bright blue (they do that when the receive information from other robots). We scored another beautiful goal with communication before the half ended and we switched to the blue team.
Now we could listen to the Game Controller, the piece of software that tells the robots what’s going on in the game (e.g. a goal was just scored, there’s 3:27 left in the half, a robot is penalized). So we could position autonomously on the field which lets us get much closer to the ball than being manually positioned (a significant advantage). As we were turning on our robots, Jayne (one of our robots) said “Head Processor Hot” which means it’s CPU is overheating. This is a very bad warning and often is a result of their fans not working. Sure enough the fan was off and we made a quick substitution: Mal for Jayne.
We again started with kick off because we were the blue team this half, though the referee forgot so we didn’t kick in to our special kick off behavior. The Game Controller sends information about who is kicking off and it sent the information too late for us to pay attention to it. Regardless, we scored again. And again. Making the score 4-0. We had some problems with our role-switching and didn’t always go after the ball or stay with the ball when we should have, but overall we looked great. We were driving toward the opponent goal and keeping it away from our own. It was a fantastic game!
We are now working on some of our minor issues for our next game at 6:00pm vs L3M. Hopefully that will go as well as this last game. We will post video of the game at some point so stay tuned!
We received all our robots back from Aldebaran this afternoon and they have remained healthy though two scrimmages. The first, against UPenn, started out strong even though we had no robot-to-robot communication. It was a fairly high scoring game but we couldn’t hold on. For the tech savvy folks out there, Wils told Brian to bind our robot socket to the broadcast address which meant we only accepted information that was coming from the broadcast address (the address that you can send digital information to and it will be forwarded to everyone connected). However, we didn’t want to receive from the broadcast address we wanted to send to it. It didn’t take long for Wils to realize his mistake and fix it for the second scrimmage later on against MRL.
This was, according to Professor Chown, the best we have looked in a very long time. First off, we were actually communicating with our teammates which allowed us to role switch (and role switch well!) for the first time in a very long time. Since our robots aren’t faster than one another, the robots dynamically decide during the game if they should be defender, offender, or chaser (the one who goes after the ball). So if the ball gets booted down-field by the current chaser, the offender may become the new chaser, the defender the new offender, and the chaser will become the new defender. Josh and Wils spent a while testing this in the lab and after finally getting communication working, it was the first time the fruits of their labor had been witnessed in Mexico. A second improvement we saw was in our localization system (knowing where we are on the field). This is rather important for positioning and during kickoff we saw some really good positioning out of our new particle filter, which relies heavily on the odometry of the robot as it’s walking while correcting with the visual information it’s seeing. You can think of it like being told to walk to your car at full speed, but getting blindfolded for 15 seconds at a time with a split second of time that you are allowed to see your surroundings before being blindfolded again. Since our motion and walk (and therefore odometry) run much faster than the cameras can take pictures, the robot is essentially walking blind for a good portion of time. But Ellis, EJ, and Octavian have worked really hard to get that to where it is and it definitely showed during the scrimmage. We also saw some improved stand-up routines and kicks courtesy of EJ, Dani, and Lizzie that were crucial in recovering from falls and scoring points. The goalie, which Dani and Lizzie have been working on, also looked awesome. It went into a defensive stance to save the ball when appropriate and even booted it across the half line from its goal box. Other kudos go to Ellis for getting separate color table functionality, Chown for getting better post recognition working as well as a color table that seemed to work really well on a field that we had never even played on before (this is pretty rare), Brian and Octavian for getting wireless connectivity to be possible, Josh for the great visual kick decision, the list goes on. We won the scrimmage 1-0 with opportunities for more goals that will be taken advantage of tomorrow morning.
We are playing NTU Robot Pal (from Taiwan) at 10:20am and L3M (joint team from France and Spain) at 6:00pm. If we place first in our pool we go on automatically to the second round robin and if we place second or third we play a intermediate game that is elimination. Hopefully we won’t go out in three strikes, but the team is very confident and we should do pretty well.
Time for some sleep and back at it in the morning!
So it’s Tuesday! Fancy that.
Yesterday we were plagued with vision issues as always. Whenever we go to a new venue, we have to take lots of pictures to calibrate our cameras and it usually takes a day or two. So after a few iterations of changing white balance, exposure, etc. we managed to see pretty well. As a result of seeing differently we have to change some of the behaviors related to the distance of things we see. For example when we chase after the ball, we see the ball slightly further away now. So we have to adjust some numbers for that to compensate for it. It’s like putting on a different pair of glasses and expecting to see the same way.
Also we’ve encountered wireless connectivity issues. This is also an issue at every competition. Somehow the organizers always buy the worst wireless routers and make it very difficult to connect to them. As a result, teams will setup their own routers, which only makes the problem worse because the interference skyrockets. But now we can at least connect to the network and communicate robot-to-robot when the network is working. But sometimes the network decides it’s just going to hate the world.
We’ve had lots of scrimmages, each better than the last (fingers crossed). We are playing games at 10:20am and 6:00pm tomorrow for our first round robin pool.
We made it to the venue to find super dark fields. The lighting is only 285 lux. If we still had the Aibos we wouldn’t be able to see anything, but lucky for us we have the amazing Lizzie, who whipped up some awesome camera parameters for us.
State of the Robots:
Aldebaran has 4 bodies and 2 heads from us.
The heads have broken fans so their processors are overheating and, if we kept using them, would melt the circuitry and be completely unusable.
We have a robot with locked knees (so they don’t bend at all) and a broken shoulder, another robot with a broken ankle, a robot with a broken neck and a broken shoulder, and a robot that won’t turn on at all due to a bad connection with it’s battery.
We are keeping our only new robot that sort of works (it’s only ailment is a tight hip) along with three new heads that are working. We also brought two old bodies with us so that we can be somewhat productive while Aldebaran is fixing our new ones.
The venue also seems to have wireless internet issues, but the ethernet has not gone down yet (fingers crossed). The organizers seem to be setting up the infrastructure (power, ethernet, wireless, etc.) as we are setting up and needing it. Small delays as compared to Istanbul when internet didn’t work at all for half the time.
The fields are not neon green and the goals look all set.
Time to get working. Pictures soon
Successful pick-up operation and our team is full strength except for the robots… More on that tomorrow
We all made it safe, scouted the convention center, and ate a pretty good dinner at the hotel. Now for the last night when we can get a decent rest