An Evaluation of Weighting Methods for Appearance-based Monte Carlo Localization using Omnidirectional Images
L. Fernandez, A. Gil, L. Paya, O. Reinoso
IEEE International Conference on Robotics and Automation  (ICRA, 2010 Workshop on Omnidirectional Robot Vision)
Ed. IEEE  ISBN:978-1-4244-5040-4

Resumen:

In this paper we deal with the problem of mobile
robot localization using omnidirectional images. We assume
that the robot is equipped with an omnidirectional camera.
In addition, we consider that the map consists of a set of
omnidirectional images with known positions in the environment.
Each omnidirectional image is represented by a single
Fourier descriptor that represents the appearance as well as
the orientation. Given an image captured with the camera at a
certain time, the Fourier descriptor allows us to find the image
in the map that is most similar in appearance. We propose the
use of Monte Carlo localization to estimate the most probable
pose of the robot. Based on these assumptions, in this paper
we propose several methods that allow to compute a weight for
each particle and carry out a comparison in terms of the error
in localization. Experimental results are presented using indoor
omnidirectional images and a real robotic platform.