Blinded by the Light: Exploiting the Deficiencies of a Laser Rangefinder for Rover Attitude Estimation
This paper presents a method to exploit inherent deficiencies in the sensing technology of a SICK laser rangefinder to detect sun positions from 3D lidar scans. Given the common use of SICK lidars on mobile robots, this method enables sun sensing for some existing configurations without requiring ad...
Saved in:
Published in | 2013 International Conference on Computer and Robot Vision pp. 144 - 150 |
---|---|
Main Authors | , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.05.2013
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | This paper presents a method to exploit inherent deficiencies in the sensing technology of a SICK laser rangefinder to detect sun positions from 3D lidar scans. Given the common use of SICK lidars on mobile robots, this method enables sun sensing for some existing configurations without requiring additional hardware or configuration costs. Adding sun sensing to mobile rovers has clear advantages; for example, sun vectors can be combined with an inclinometer to calculate rover orientation in an absolute reference frame and used to improve pose estimates. The proposed sun sensing technique was verified using a SICK LMS-511 lidar mounted on a Schunk panning unit through two separate experiments. In the first experiment, the outputs of both our algorithm and a Sinclair Interplanetary SS411 digital sun sensor were compared to solar ephemeris data over an entire day. While the SS-411 has higher accuracy, the experiment showed that our lidar-based method has acceptable accuracy and a larger field of view (FOV) that covers the entire sky. In the second experiment, our sun sensing algorithm was used with an inclinometer to calculate the absolute orientation of the rover periodically during a traverse. This information was used with wheel odometry to estimate rover poses over the entire traverse, yielding more accurate results than wheel odometry alone. When including lidar-based sun measurements, the average estimate error over the entire traverse was only 8.4 metres, an 88% improvement over wheel odometry (70.4 metres). The resulting final position estimate error was 22.8 metres, or 2.76% of total distance travelled. |
---|---|
ISBN: | 1467364096 9781467364096 |
DOI: | 10.1109/CRV.2013.37 |