Development of a fast self-localization algorithm based on laser range finders

摘要

This paper describes a new, fast self-localization algorithm based on laser range finders for use in the Intelligent Ground Vehicle Competition (IGVC) Auto-Nav Challenge. This challenge course uses circular cone shapes as obstacles; thus, we utilize this shape to achieve fast, real-time self-localization. To detect the accurate positions of circular cone obstacles, regardless of the robot's observed direction, we apply the circular Hough transform. To robustly estimate the mobile robot's posture, we formulate equations between the geometric relation and the traveling direction, and then solve these equations by applying singular value decomposition. To estimate fast and stable self-localization, we fuse the robot's estimated posture and absolute position from GPS by applying a complex-type Kalman filter. The validity of the proposed algorithm is confirmed using actual outdoor environments.