Extraction of orientation from floor structure for odometry correction in mobile robotics

Abstract

We are presenting a method for correcting odometry readings of a robot for increased accuracy of position estimation. Our method uses a simple pragmatic approach and exploits the distinct structure of the floor in our experimental area. By continually extracting orientation information from the floor view, we are able to correct the heading component of odometry, thereby eliminating the major source for position errors. Compared to other approaches the solution is computationally inexpensive. Our experiments show that by employing our correction method we are able to significantly increase position accuracy and consistently map paths up to several hundred meters. © Springer-Verlag Berlin Heidelberg 2003.