Abstract
This paper describes a laser range finder based new fast self-localization algorithm for IGVC Auto-Nav challenge. In order to achieve fast real-time self-localization, we utilize shape of circular cone which exists as obstacles around the IGVC Auto-Nav challenge course. To detect accurate circular cone obstacle position regardless of observed direction of mobile robot, we apply circular Hough transform to detect center position of circular cone obstacles. In order to estimate mobile robot posture robustly, we formulate equations between geometric relation and traveling direction and to solve by applying singular value decomposition. To estimate fast and stable self-localization, we fuse between estimated mobile robot posture and absolute position from GPS by applying Complex type Kalman filter. Validity of proposed algorithm is confirmed by actual outdoor environments.
