Section: New Results

Reliability estimation and information redundancy for accurate localization

Participants : Zayed Alsayed, Anne Verroust-Blondet, Fawzi Nashashibi.

Our goal is to improve localization systems performances in order to be able to navigate in urban and peri-urban environments. For this purpose, we choose to study the reliability of a SLAM method that incrementally builds a map of the surrounding environment from an information given by a set of 2D laser points.

This year, we focused on SLAM failure and non-failure scenarios.

• Experimental data acquired on the VEDECOM demonstrator in the context of ITS Bordeaux demonstrations in 2015 were analyzed. This evaluation showed in [30] that the SLAM concept seems better suited to urban scenarios, while algorithms such as lane marking detection could offer a good alternative in peri-urban environments.

• In parallel, we worked on designing a reliability measure associated to the pose given by our SLAM considering the geometrical configurations of the 2D laser points describing the environment and the computations done in the maximum likelihood matching process.