Robot Localization by Stochastic Vision Based Device

The localization of a mobile vehicle in an dynamic outdoor environment is a very hard task that requires robust data processing and interpretation. This paper proposes a localization device, based on both image processing and stochastic evaluation. As far it is concerned the data processing task is performed by a vision device that uses a principal component analysis in order to find out the most probable position of the mobile vehicle considering the image acquired. This vision device works in parallel with a stochastic position evaluation that uses Partially Observable Markov Decision Process where the observation probability is conditioned by the results of the vision device. The method developed in this paper seems to give very good results compared with the standard methods and particularly the neighborhoods based ones.