Hierarchical Image-based Localisation for Mobile Robots with Monte-Carlo Localisation

Abstract

This paper extends our previous works on image-based localisation for mobile robot. The image-based localisation consists in matching the current view experienced by the robot with the reference views stored in the visual memory of the robot. The original idea was to use the Fourier components as signatures for the omnidirectional images acquired by the robot. The extensions proposed in this paper are: the possibility to have a localisation with different accuracy while the robot navigates (called hierarchical localisation) and the introduction of a Monte-Carlo localisation technique to increase the robustness of the system in environments where the image-based localisation can be mislead. Experiments demonstrated the feasibility of the hierarchical localisation and the robustness of the implementation of our Monte-Carlo localisation.