Section: New Results

SLAM failure scenario detection for laser-based SLAM methods

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

Computing a reliable and accurate pose for a vehicle in any situation is one of the challenges for Simultaneous Localization And Mapping methods (SLAM) methods [18]. This year, we worked on the detection of SLAM failure and non-failure scenarios and a technique detecting a priori potential failure scenarios for 2D laser-based SLAM methods has been introduced. Our approach is independent of the underlying SLAM implementation as it uses raw sensor data to extract a relevant scene descriptor, which is used in a decision-making process to detect failure scenarios. Experimental evaluations on three realistic experiments show the relevance of our approach. See [22] for more detail.