The "soccer moms" at Cornell University don't drive mini-vans or call out "that was off-sides" while the kids kick the ball around on the field. Instead, they communicate with their soccer playing robots over a wireless connection.



Cornell University students have entered their robot prototypes in the RoboCup competition every year since 1999. They won the world championship four times. The competition helps the student team learn about artificial intelligence, distributed computing, program verification, and this year, embedded systems that allow the robots to be autonomous.

RoboCup is an international joint project to promote artificial intelligence, robotics, and related fields. The event fosters research by providing a standard problem - win the soccer game -- and integrates a wide range of technologies. The students participating are mechanical engineering, electrical engineering, and computer science majors. Several PhD students are also on the team.

The challenges associated with developing an autonomous robot team that can compete in a soccer match are substantial. The Cornell team decided to treat this year's competition as a research year. They knew that the modifications they needed to make might set them back a bit.

Aided by an Innovation in Excellence award from Microsoft Research, the team put more intelligence into the robots rather than depending on a workstation off the field. This puts them in a better position for future events, when local vision will become mandatory. Robots with local vision have the camera mounted on the robot, those with global vision 'see' from a camera mounted on top of the field. The on-board intelligence also allows them to improve motion control accuracy, decrease system latency, and explore distributed decision-making strategies.

"I definitely think this was a great learning experience. It's a hands-on way for students to apply the things they learn in class. Also, working on a team with students from several disciplines gives the students an appreciation of how complicated the interfaces between the different fields can be. They also found out how much things can go wrong when people don't communicate properly," said Professor Raffaello D'Andrea, the faculty advisor for the project and a leading expert in real-time control of autonomous systems.

The robots are omni-directional and have four wheels. Their new operating system, Windows CE, and the PC/104 form-factor computer gives them increased computing power in the compact space available to meet competition regulations. The soccer field is a platform that looks something like an over-sized ping-pong table.

"Our biggest challenge this year was to move strategy and control from a workstation computer into the robots," said D'Andrea.

"In addition to the distributed artificial intelligence we also had to develop a new wireless communication sub-system for debugging. In previous years the strategic decisions were made on a computer off the field and broadcast to the robots," explained D'Andrea.

They also used a different drive train to move the robots around the field. By using brushless instead of brushed motors, they gained a higher power density but found that the robots were more difficult to control. They had to design drive electronics for the motors.

(Click here for enlarged photo of robot internals.)

"The robots only play soccer, but given different sensors and actuators, the majority of the work is transferable to 'real-world' automation challenges," said D'Andrea.

"Our main goal this year was to contribute to the RoboCup by pushing the envelope further and bringing new technology to the league. We succeeded in showing that on-board processing doesn't have to reduce performance. On the other hand, we weren't able to bring a fully working system to the competition, due to the increased complexity. We lost one game, won one, and tied three. Compared with our record, that wasn't satisfactory. But we look forward to next year's competition."

Planned refinements for next year include the use of 802.11a as a replacement for this year's Bluetooth wireless system, and integration of local vision into the robots.

Stewart Tansley, a program manager for University Relations at Microsoft Research, is excited about the direction that the Cornell team has taken.

"PC technologies are starting to transform robotics, enabling much more powerful software than in previously pervasive microcontroller based devices, and Cornell's excellent robotics work is pushing the envelope in all the right directions," said Tansley.

Visit the Cornell RoboCup project, here.

---

Summary of system specs:

---

• Based on Advantech PCM3360 PC/104 single-board computer (photo):
  • 400 MHz Intel Celeron processor
    
  • 128MB SDRAM,
    
  • 16MB CompactFlash storage
• Windows CE 4.2 embedded OS
  
• Software written in C++
  
• Custom-designed digital input/output (photo):
  • 5 channel motor control
    
  • 1 channel solenoid kicker output
    
  • Sensor inputs for rate gyro, two accelerometers, infrared beam, motor encoders
    
  • Navigation sensors (Encoders, Rate Gyro, Accelerometers)
    
  • Ball detection sensor
• Wireless communications:
  • RF modem for transmitting position and orientation data
    
  • Bluetooth for inter-robot communication
• Camera
  
• Omni-directional drive
  
• Solenoid kicker

---

Members of the Cornell student team:

---

• Jin-Woo Lee
  
• Oliver Purwin
  
• Mike Sherback
  
• Patrick Dingle
  
• Jieun Kim
  
• Jim Mondello
  
• Joe Golden
  
• Nathan Pagel
  
• Aaron Nathan
  
• Eugene Byrne
  
• Ivan Han Hor Siu
  
• Sergei Lupashin

---

Copyright © 2003-2004 Microsoft Corp. All rights reserved. This article was initially published on the Microsoft Research website. Reproduced by WindowsForDevices.com with permission.

Discuss Robotic Soccer Moms also do Windows (CE)   >>> Be the FIRST to comment on this article!

>>> More Windows For Devices Articles Articles          >>> More By Doug