Direction augmented probabilistic scan matching

Abstract

The scan matching is widely used for localization and mapping of mobile robots. In this paper, a direction of data point in the scan is approximated and this is incorporated into the scan matching algorithm to improve the performance. The direction of data point is the normal direction of the least squares fitted line based on neighbors of the data point. Owing to this incorporation, the performance of the scan matching can be improved. The number of iterations decreases, and the tolerance against a large relative-rotation between scans increases. Using real sensor data of a laser range finder, experimental results verify the performance of the proposed algorithm, the direction augmented probabilistic scan matching.