Sample-based techniques for state estimation in RoboCup

by
Harr Chen

One of the major components of soccer-playing robotic dogs is the ball tracking system. The robot needs to be able to accurately and reliably estimate the position and velocity of the ball in spite of noisy observations. In this talk I will present the extended Kalman filter (EKF), a way of integrating prior estimates and current observations to obtain an accurate picture of the ball's movement. I will also discuss ways in which the EKF can be improved to work well with rapid significant changes in the ball state (when the ball is kicked, for example), and ways in which the EKF can be combined with other techniques to perform ball tracking in tighter conjunction with state localization.

Advised by Dieter Fox

MGH 228
February 24, 2003
3:30 - 4:20 pm