Experiments on Augmenting Condensation for Mobile Robot Localization

Patric Jensfelt, Olle Wijk, David Austin, and Magnus Andersson

Abstract:

In this paper we study some modifications of the, by now well known CONDENSATION algorithm. The case studied is feature based mobile robot localization in a large scale environment. The required sample set size for making the CONDENSATION algorithm converge properly can in many cases be too high with respect to computational complexity. This is often the case in area with lots of (feature) symmetries, for instance long corridors. To deal with this problem we experimentally study two small modifications of the CONDENSATION algorithm. The first strategy called ``CONDENSATION with random sampling'' is to take part of the sample set and spread it randomly over the environment the robot operates in. The second strategy called ``CONDENSATION with planned sampling'' is to place part of the sample set at planned posititions based on the features detected. From the experiments we conclude that it is possible to manage with a sample set size which would be insufficient for the original CONDENSATION algorithm.

BibTeX Entry:

@InProceedings{Jensfelt00b,
  author = 	 {Patric Jensfelt and Olle Wijk and David Austin and Magnus Andersson},
  title = 	 {Experiments on Augmenting Condensation for Mobile Robot Localization},
  booktitle = 	 {IEEE Intl. Conf. on Robotics and Automation},
  year =	 2000,
  pages =        "2518--2524"
}

Download: ps.gz (101k) or pdf (269k)