Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
SAF Archives  search ask the scientists in the classroom cool science
scientists from previous shows
cool careers in science
ask the scientists
PREVIOUS SCIENTISTS

Photo of Manuela M. Veloso Manuela M. Veloso as seen on Natural Born Robots: Go, Team!

Click on Manuela's photo to read a brief bio.



q Have you thought about making different robots for specific tasks? For example, making a bigger robot for goalie? Or faster robots for offensive moves? (Question sent by Adam)

A The rules of the game impose strict maximum sizes for the robots, so we cannot make the robot goalie any bigger... In fact, the goalie robot is already of a different size than the other robots within the maximum allowed size, namely it occupies the maximum surface (approx. 150cm2) and its largest dimension (base diagonal) is 18cm (see http://www.robocup.org/ for the complete set of rules; they are now under slight revision; the new set of rules will come out soon).

As for faster robots, that's one of our goals! Notice however that the speed of the robots is a function of how fast our vision algorithm can process images. Think about the following: our vision processes now 30frames/sec (we are working on improving it to 60frames/sec). This means that only 30 images are captured per second. If a robot (or ball) moves at a speed of 2m/s (fast, but not very fast) then the camera sees the moving object about every 7cm. For faster robots (or ball) we need to be able to process images faster. We are working on better hardware for the robots to allow for that and faster vision! :-) One final comment, our attacking robots deliberately move faster now when shooting into the goal than when moving around obstacles.



q I was captivated by the performance of the soccer robots. The show centered generally on their amazing group performance, but I was equally interested in the technical details of their construction. Is there any material available (URL?) that describes their physical construction (types of motors, batteries, sensors), hardware (embedded controller, PC-104), operating system (linux, QNX) and means of communication? Good luck in future matches! (Question sent by Gary Huntress)

A Thank you for your nice words and for your wishes! We have several technical papers about the team - see the list of publications available at http://www.cs.cmu.edu/~mmv, including the robots built in 1997. These research papers focus mostly on the intelligence and strategies of the robots. Unfortunately we don't have yet a complete description of the hardware. Probably because we are always trying to improve the motors, batteries, sensors, radios, etc, so we don't think that the current version of the robots is the final best one... We are planning on adding soon a list of the parts used on the CMUnited-98 robots to our Web page. So keep in touch!



q Will there be a robot soccer game on the market any time soon? (Asked by many viewers)

A There is a simulation environment, namely the RoboCup server, that is available free -- you can download it from the RoboCup Web site, at http://www.robocup.org/. Note that the playing agents need to be programmed by the users. Examples of source code are available off our robotic soccer Web page, at http://www.cs.cmu.edu/~robosoccer/ -- follow the links to CMUnited-98 simulation.



q I would like to build a simple robot. If you have any suggestions, or know any websites or books with directions I would really appreciate it. (Asked by many viewers)

A Unfortunately this is not a well-documented process... I am not myself an expert robot builder, although I supervised the building of the robots in the CMUnited teams. And both CMUnited-97 and CMUnited-98 were the RoboCup-97 and RoboCup-98 world champions. My main research focus is on the development of the "intelligence" (e.g., memory, planning, learning) of the robots.

But in very simple terms, here is what you need to know: A simple robot consists of three main components: mechanical, electrical, and control. For the mechanical part, you need to build the robot (chassis) and its locomotion capabilities. You need to search for motors, gears, encoders, motor controllers, etc. For the electrical part, you need to worry about the power supplies, and all the connections between the intelligent control and the motors. Some sensors will need to be added (e.g. vision camera, radio receiver, or acoustic/touch sensors) and possibly actuators (e.g., a kicker, a gripper). The control is achieved by the amount of computation that you can afford - it can be a simple microprocessor or a full computer. Think of it as building a small electrical car and then making it intelligent! Hence, you can use catalogs and Web sites for hobby toy stores/mechanical parts/etc. The main challenge, I think, is to make it autonomous, i.e., to be capable of writing a program that makes the robot use its sensors to perceive the world and make intelligent decisions to achieve its goals.


q My class watched "Natural Born Robots" and asked a question about the soccer game. Why did the Korean Team only have three players and FuFighters have five or six? (Question sent by Hennessey-Liptrot 7th Grade Science Teacher North Cumberland Middle School)

A The number of robots is specified by the rules of the RoboCup small-size league (see http://www.robocup.org/). Currently, there can be up to five robots and no more than five. FuFighters had 5 robots. Korea chose to have only 3. That's fine. The number of robots allowed and their maximum size is a function of the size of the playing field, so the field is not too crowded. But we don't want to encourage teams with just a few players (imagine just one player!) because the main research question underlying RoboCup is the study of systems of multiple robots.

q Why did you choose to work on soccer robots instead of basketball robots or football robots? Are you and your colleagues ever going to make robots that play other sports? (Question asked by many viewers)

A My Ph.D. student Peter Stone and I were inspired by a small group of researchers who, as far back as 1993 (e.g., A. Mackworth, H. Kitano), had the idea of robotic soccer. RoboCup, as the international robotic soccer association (see http://www.robocup.org/), took over very rapidly and we have been involved in doing research in robotic soccer for several years. I participate with several of my students in three different leagues: simulation; small wheeled robots (as seen on the Natural Born Robots episode of Scientific American Frontiers); and the Sony legged robots, as also seen on this episode. Robotic soccer seemed a first step in robotics sports, as the ball could be pushed into a goal by a little moving wheeled robot. Compare this with being able to throw the ball into a basket! But there are actually big robotic arms that are capable of exactly throwing a ball into a basket! However no complete robotic basketball game was done yet. Other sports are for sure on the way for robots to play. But, as of now, we are concentrating on robotic soccer until we can get a team of full humanoid robots playing against human soccer players!




 

Scientific American Frontiers
Fall 1990 to Spring 2000
Sponsored by GTE Corporation,
now a part of Verizon Communications Inc.