Robotics and Soccer with Ashley Stroupe

 

Background

Ashley Stroupe is a fourth year PhD student, studying robotics at Carnegie Mellon University. In this interview, she shares her experience at the International Robocup Competition that she participated in over the summer as well as her current academic and research goals in the field of robotics.

 

Questions and Answers

What did you do over the summer?

Ashley: Over the summer, I worked on the Mid-size Robocup Competition, which is an international Robotics Soccer competition.  In mid-size, you compete with robots that are about the size of a small dog.

 

So how did your team do at the competition?

Ashley: Our robotsí programs were very competitive in terms of the abilities that our robots had. They were sensing the environment very well. They were seeing the ball very well. They were doing intelligent things. But the robots that we were using were just much slower than other teams, so even though they knew what they needed to do, they just could not get there quickly enough. So, we lost in the qualifying round and did not make it on to the playoffs, like we did last year. In terms of the software aspects, which is what most of us on the team were involved with, we were pretty pleased with how it worked.

 

When was the Robocup project first started?

Ashley: The first Robocup was 10 years ago. But theyíve added events, and gotten many more teams to compete, and have really grown over the years. This year there is a simulator team, which is mostly like human soccer and is all done on a computer screen. There is the small-size, mid-size, and the Sony dogs. They are also going to be starting a humanoid league, which are little walking robots (about a foot and a half tall), that right now, sometimes fall over when they try to kick the ball.

 

Do you use a real soccer ball in the competitions?

Ashley: It depends on the size of the robot. The mid-size league uses a real soccer ball, painted orange so they can see it better. The smaller robots use something the size of a golf ball and the pets use something about the size of a tennis ball. They scale it with the robots, just like they scale the field.

 

How long does the competition normally last?

Ashley: Each game is two ten minute halves. So itís fairly short, but also they have a lot of games to get through. In the small-size league, they stop the clock every time the ball goes out of bounds. In the other leagues, they just run straight through so the games go much more quickly. First, they have a round robin run that gives each team a seeding, and the top seeded teams are chosen to compete in the qualifiers. The whole competition runs about a little over a week. Because you have setup and practice days where everyone goes in to calibrate their cameras, and fix anything that was broken during shipping. Then you have a couple of days of competition. They interweave the competition days a little bit so people will have time to watch some of the other games going on. And then there are two days of workshops, where people present papers on interesting ways theyíve solved a problem that might be useful to other people.

 

Have you ever had the problem where the robot scored the wrong goal?

Ashley: That does happen. This was a much less of a problem this year than it was last year. Occasionally, just because the robots are not terribly agile, when they are trying to get behind the ball to knock it out of the way, they can bump it because they are closer to it then they think they are and knock it in the goal. But again, thatís getting much better.

 

How far off is Robocup soccer from real soccer? What advantages do human players still have?

Ashley: Speed is one. Also, manipulation. The human feet are really amazing for handling a soccer ball, and the robots are certainly nowhere near as graceful or agile. Also, sensing. We can look around the room, and without even thinking about it, we can immediately recognize a person regardless of what posture they are sitting or standing in. We recognize tables and chairs.  Our vision processing is very elaborate and complex. Robots cannot do that. We have to tell robots how to interpret things, and that makes programming robots very difficult.

 

One of the problems I noticed in a video of with the robot dogs is that they can tip over. Do mid-size robots have that problem too?

Ashley: No, the ones that we use are wheeled robots. So they are very stable. Our robots are very short compared to a lot of the other robots, just because we bought them and thatís how they are; they are also very lightweight. Last year, one of our robots got caught on somebodyís kicker, and as they pushed further into us, they actually did flip us over. It also crashed our computer too, but that almost never happens.

 

Whom do you compete with?

Ashley: Most of the competitors are from Europe. Europe is really into soccer. Since that is their favorite sport, doing robotic soccer is bigger to them then it is in the US. So there are several teams from Germany and Italy, plus others from all over Europe. Most of them are affiliated with universities or private research labs.

 

Do most teams build their own robots?

Ashley: In the mid-size, itís about half and half. A lot of people take some kind of commercial robot and adapt; they write the software and add a kicker, etc. The other half of the teams build them from scratch.

 

What kind of commercial robots did your team use?

Ashley: We use the Cye-Robots, which is made by a company here in Pittsburgh. It was designed to pull a vacuum cleaner around your house. It has two wheels and a trailer, so it can turn and drive in any direction. On the trailer, we attached the laptop that talks to the robot. The robot itself has a very primitive controller such that all it can do is drive around and bump into things. But we do all the higher-lever processing on the laptop and send it to the robot. Then the robot interprets it, and drives in an appropriate direction.   

 

How did you become interested in Robotics?

Ashley: I guess Iíve always been interested in different kinds of engineering. My dad was an engineer and got me interested in this stuff when I was very young Ė especially the space program. These days, our space program is pretty much a robotics program. And since I wanted to work for NASA, I thought robotics would be a great thing to do. So I began studying robotics, and I really enjoyed it. Itís a great subject because it incorporates all different aspects of science and engineering, such as physics, mechanical structures, and software.

 

What are you currently working on, in regards to your Robotics degree at CMU?

Ashley: Well Ė I will be proposing for my thesis in December.  I will be working on some aspects of robot teamwork, mostly dealing with what we call cooperative localization. One of the biggest problems in robotics is that when you have a robot driving around, it needs to know where it is in order to accomplish any tasks. This has been reasonably well solved for one robot driving around looking at its environment, and figuring out where it is. But when you have multiple robots trying to work together, the problem becomes more complex. It is not good enough for each robot to maintain its own model of the world, because if that doesnít match with what the other robots think, they wonít be able to work together very well. And so its dealing with that issue, not only building up these models but sharing them and making sure that all the robots who work together have the same model, and the best possible model; that is what I will be working on. 

 

How exactly do the robots determine where they are?

Ashley: There are lots of ways of doing it. The most common way is either by sending the robot a map ahead of time, or by telling the robot where specific things are in the environment that it can recognize. It will then observe these things (either through a camera, sonar, or laser) and get a distance and an angle to that landmark. And you combine that with the model that you have of the robot (for example, the robot wants to drive forward a meter and it has some model of how close to a meter it actually gets). In other words, you combine what your sensors tell you, with what your wheels tell you (called odometry), and get some estimate as to where youíre actually are thatís better than what you would get with either one alone.

 

How is the PhD program at CMU?

Ashley: The PhD program I think is great. Like any other PhD program, it is certainly a lot of work. Itís about five and a half years for most people; we have about two and a half years of course work, followed by research. The CMU PhD program provides a great community. The people are really supportive. We have people that do every aspect of robotics, which is pretty unusual. Nowhere else has the kind of resources in terms of people available as CMU does. So whatever problem youíre working, there are always three other people working on the same problem that can help you if you get stuck. There is very little bad competition. Everybody is very supportive and helpful.

 

Are there are lot of women in the PhD program?

Ashley: It goes up and down. In my year, which was a total fluke, we had 18 PhD students and six of them were women. Since then, weíve had 1 or 2 women per class and before that, we had one, one or zero women. For me, I think I was just lucky. I was in a class that had a lot of women. I went to an undergraduate school that had enough women, but we were never very close. It was more competitive; we all felt like we had something to prove and so we didnít form any tight bonds. Here, it has been amazing. Weíre very close, hang out together all the time, and are supportive of each other.

 

Why did you choose to come to CMU for a PhD in Robotics?

Ashley: CMU is the only place that you can actually get a degree in robotics. Everywhere else, itís a degree in computer science or something else. So CMU seemed to be the obvious place. Not to mention, MIT does good research, Stanford does good research, Maryland does good research, but they all only have 1 or 2 big projects, and maybe 5 really active faculty members in robotics. Here, there are 60 faculty members. There is just no comparison in terms of the environment for the richness of what you have here. If you can think of anything that has to do with Robotics, you can find somebody working on it here.