CM-Trio-99

1 Overview The robots used in this competition were generously provided by Sony 3]. The robots are the same as the commercial AIBO robots except for slight hardware changes and programming capabilities. These autonomous robots are about 30cm long and have 18 degrees of freedom. The neck pans 90 allowing the robot to scan the eld with its on board camera. Six uniquely colored landmarks are placed around the eld (at the corners and center-line) to help the robots localize. Each team consists of three robots. Like our team last year, CMTrio-98 5], we divided our team between two identical attackers and one goalie. We divided our system into three main components: vision processing, local-ization, and behaviors. The vision system is responsible for calculating distance, angle, and conndence measures for all objects visible by the robot. The localiza-tion is responsible for calculating the position of the robot on the eld given the movements executed and the landmarks seen (we did not implement goals for localization this year). The behaviors are responsible for taking this information and winning the game. Results from our matches in RoboCup-99 at Stockholm show our algorithms to be eeective. Our team won all but one of its games, and the one it lost was lost by only one goal. Our team was the only one in this year's league to score goals against opposing teams and never to score a goal against itself. Our goaltender was the only one in this year's league to score a goal itself. 2 Vision The vision system processes images captured by the robot's camera to report the locations of the ball, the 6 unique location markers, the two goals, and the robots. The main steps in vision processing are: 1) capture an image and classify each pixel's color in hardware using predetermined color thresholds, 2) nd connected regions of the same color, 3) merge close regions of the same color, 4) use geometric lters to remove false positives, 5) calculate distance and angle of objects in ego-centric coordinates. The on-board camera provides 88x60 images in the YUV space at about 15Hz. Hardware Color classiication is then performed on these images. The thresholds for color segmentation are created by a supervised learning method based upon hand labelled images captured from the dog's camera. This results in a new image indicating color class membership rather than raw camera colors. This image is …