Robust Mobile Robot Localization by Tracking Natural Landmarks

This article presents a feature-based localization framework to use with conventional 2D laser rangefinder. The system is based on the Unscented Kalman Filter (UKF) approach, which can reduce the errors in the calculation of the robot's position and orientation. The framework consists of two main parts: feature extraction and multi-sensor fusing localization. The novelty of this system is that a new segmentation algorithm based-on the micro-tangent line (MTL) is introduced. Features, such as lines, corners and curves, can be characterized from the segments. For each landmark, the geometrical parameters are provided with statistical information, which are used in the subsequent matching phase, together with a priori map, so as to get an optimal estimate of the robot pose. Experimental results show that the proposed localization method is efficient in office-like environment.