Optimizing Precision of Self-Localization in the Simulated Robotics Soccer

We show that all published visual data processing methods for the simulated robotic soccer so far were not utilizing all available information, because they were mainly based on heuristic considerations. On the other hand, we show that the improved accuracy could result in a significant score gain even for a reasonably good simulated team. Some researchers have approached to estimating the agent location and orientation as two separate tasks. This had caused systematic errors in the angular measurements. Further attempts to get rid of them (e.g. by completely neglecting the angular data) only aggravated the problem and resulted in the losses in he accuracy. We treat the estimation problem in a rigorous way, by jointly estimating the agent Cartesian coordinates and its view direction angle using the extended Kalman filtering technique. The experiments with the proposed method run in the simulated soccer setting give the idea of the achievable average error limit for this particular appl...

• Type: Conference
• Year: 2004
• Where: ROBOCUP
• Authors: Vadim Kyrylov, David Brokenshire, Eddie Hou
• Keywords: Data Processing Methods, ROBOCUP 2004, Significant Score Gain, Simulated Robotic Soccer