Séminaire de l'équipe DISCO :

Improving the Integrity of Unmanned Ground Vehicle Perception Systems.

Le 10 Juin 2010 à 11h00

Intervenant

Christopher Brunner
Australian Centre for Field Robotics (ACFR), The University of Sydney


Lieu

LAAS-CNRS - Salle Europe

7 avenue du Colonel Roche

31077 TOULOUSE Cedex 4

Résumé

Unmanned Ground Vehicles (UGVs) require high integrity perception systems in order to achieve long-term autonomous operation. Current state-of-the-art perception modules employ a range of sensors along with a variety of algorithms to interpret and fuse data to provide a meaningful Representation of the Environment that the UGV uses to navigate autonomously. Despite significant effort in the robotics community over the last few decades, most failures in UGV systems still occur due to perception problems. The most common sources of failure are not from sensor faults but rather from a failure to interpret the perception data correctly for the application. This seminar discusses the use of quality metrics and non-direct redundancy to evaluate multimodal sensor data and detect failures in perception systems. This is illustrated with Visual and Infrared cameras in the presence of challenging environmental conditions (such as airborne dust and smoke) to improve the reliability of UGV perception systems.





Retour

Export au format PDF	Ajouter à mon calendrier