Robot soccer is a growing arena for the development and advancement of robotics algorithms and research. Many of the universities building robot soccer teams use large groups of graduate and undergraduate students. Bowdoin is developing a team, led by Greydon, with the goal of being competitive on an international scale despite not having any graduate students.
The 4-legged Robocup league consists of teams of four Sony Aibo robots. Aibos have onboard cameras and a number of other sensors to aid their playing skills. Getting the Aibos to play soccer requires the development of algorithms in four major areas: 1) vision, 2) motion, 3) localization, and 4) behavior selection. The vision system must be capable of quickly recognizing key landmarks on the field (e.g. the opponents' goal) as well other players. The motion system must be capable of moving the dog as quickly and efficiently around the field as is possible, as well as executing specialized behaviors suck as kicks. The localization system takes input from the vision system and uses it to determine where on the field the dog is. Finally, the behavior system determines how the dog should act. For example, it will need to make decisions on whether to pursue the ball or to fall back into a supporting position for its teammates.
The goal of Greydon's research is to have Bowdoin ready to field a team in the 2005 American Open in May. The team will be built on a combination of work done in Bowdoin's Robotics course, ideas from other robocup teams and Greydon's own research.