Robot Self-localization with Optimized Error Minimizing for Soccer Contest

Abstract

This paper proposed an integrated solution for robot self-localization in the RoboCup middle size league. Firstly, a kind of radial scan-lines are introduced to preprocess the images from the omni-directional camera so as to reduce the computational load and improve efficiency. The key points of this algorithm are minimizing the errors between the mark points and the detect points. In order to improve the self-location performance and reduce the effect by the dithering image, an optimized approach was presented and the extended Kalman filter was utilized to fuse visual information, odometry information and digital compass data. Secondly, before applying this approach into our robots, we built the simulate environment and compared the result to the existing particle filtering approach with 200 particles. Simulation result showed this approach is computational efficient and precise. Our method is computationally cheaper than the particle filtering approach. Finally, we tested the approach in the real RoboCup middle size league robot contest field. The experiments showed the optimized error minimizing self-localization provided high level capabilities.