Positioning of the mobile robot LiAS with line segments extracted from 2D range finder data using total least squares.

In this paper, a new algorithm is described for dynamic map building with geometrical primitives for a mobile robot. The dynamic map is built up using a 2D range finder mounted on the mobile robot LiAS1 which is navigating in the environment. The dynamic map is used for both planning and localisation purposes. The map is composed of line segments and circles. For the extraction of line segments, a total least squares algorithm is used. The parameters describing the geometrical primitives are provided with uncertainties which are used in the matching phase and which are necessary if the map is used for localisation. This paper describes in detail how the uncertainty on the robot position and the uncertainty on a single range measurement leads to the uncertainty on the parameters of a geometrical primitive. Promising experimental results obtained by the algorithm in real unstructured environments are presented. [ABSTRACT FROM AUTHOR]