A novel localization method based on sharp corners for indoor mobile robots

Abstract

For a conventional Corner Matching Method (CMM), a corner is identified only when two edges of the corner are sensed (scanned). Utilizing the geometric features of sharp corners, a novel and general CMM method is proposed in this paper. The major advantage of this approach over the conventional CMM is that a corner with only one sensed edge can be identified through comparing the scan data between two subsequent sensed points in the same scan, as a sharp corner with only one edge sensed will cause two adjacent scan data an abrupt change. In order to identify the same corner point in two adjacent scans, an effective corner matching algorithm is formulated based on the distance matrix constructed from the corner points identified in each scan. Finally, with two or more matched corner points between two adjacent scans, the change of the robot pose (position and orientation) can be accurately estimated. As the proposed approach only deals with the point data, it is computationally efficient. The effectiveness of this approach is validated through experiments.