Posted June 08, 2007
Since 1999, the Gibbons Summer Research Program has been funding students and faculty to collaborate on projects using technology. John Gibbons '64 established the fund to enable rising sophomores, juniors, and seniors to work with faculty members on projects that will enhance their academic progress and explore interdisciplinary issues. As a result of Mr. Gibbons' generosity, the program has funded over 60 students.
During the summer of 2009, the Gibbons program funded ten students including, Elise Krob '10 who worked on the RoboCup project with Eric Chown, Samuel S. Butcher Associate Professor in the Natural Sciences. RoboCup is an international organization that promotes research, collaboration, and code development in the fields of artificial intelligence and cognitive science through the medium of robot soccer. In 2007, Bowdoin's RoboCup team, the Northern Bites, won the World Championship Games held in Atlanta, GA and this past summer they finished second in Graz, Austria. Bowdoin hosted the RoboCup U.S. Open at the Sidney J. Watson Arena In May and the Northern Bites finished in fourth place.
Until recently the standard robot was a Sony AIBO dog, which Professor Chown and his RoboCup team used when they won the World Championship in 2007. However, the goal of RoboCup is to develop a team of autonomous robots that can defeat humans by 2050 so the new robot, the NAO robot by Aldebaran, has two legs instead of four. The new robots presented Elise and Professor Chown with new challenges and obstacles and opportunities to update their code base.
One of the biggest differences between the two robots is their vision capability. The new NAO robots have a much better camera than their four-legged predecessors and can more clearly "see" objects in the form of pixels in front of them. As a result, more pixels need to be processed. Elise and Professor Chown used "blobbing" and "run-length encoding" to take streaming images from the camera and scan every pixel to assign colors and values to different objects the robot sees. Elise explained that, "blobbing is the idea that if a group of pixels of the same color are grouped together, they probably belong to the same object, i.e. the ball. Run-length encoding is a way to compress data by storing connected pixels of the same color together, for example the field." This is just one way to create blobs.
The team hopes that because the NAO robots view similar images repeatedly during a match that they will develop a cognitive vision system that keeps a memory of the image just seen and can anticipate the next image and action. This would allow them to react faster to what others are doing and to set expectations for what is happening on the field. Elise explained that, "Vision is crucial to the operation of the robots, and an efficient and accurate vision system would give us more time to execute behaviors, communicate among team members, and hopefully score some goals. Being able to see is the first step in building all other behavioral, locomotive, and communicative code, and making it a more cognitive process will lead to it becoming a better system."
Elise is a psychology major with an interest in both cognitive psychology and cognitive architecture. "The marriage of computer science and psychology in cognition nicely mirrors my academic interests and provides me with rich and interesting research opportunities in the field of cognitive science."
Professor Chown, Elise, and the rest of the RoboCup team will host a match during Parents' Weekend and hope to host the RoboCup U.S. Open again in the near future.
For more information about the Gibbons fellow ship and projects please go to our Gibbons Summer Research Program site.
"Vision is crucial to the operation of the robots, and an efficient and accurate vision system would give us more time to execute behaviors, communicate among team members, and hopefully score some goals."
— Elise Krob '10