Stéphanie Lefèvre - Risk estimation at road intersections for connected vehicle safety applications

Organized by: 

Stéphanie Lefèvre


Stéphanie Lefèvre

J’ai le plaisir de vous inviter à ma soutenance de thèse intitulée "Risk estimation at road intersections for connected vehicle safety applications", préparée au sein de l’équipe e-Motion et encadrée par Christian Laugier (Directeur de thèse) et Javier Ibanez-Guzman (superviseur Renault). La soutenance aura lieu le lundi 22 octobre 2012 à 10h dans le Grand Amphithéâtre de l’Inria à Montbonnot, et se déroulera en anglais.

Jury :

  • Roland Chapuis, Université Blaise Pascal Clermont-Ferrand, rapporteur
  • Christoph Stiller, Karlsruher Institut für Technologie, rapporteur
  • Urbano Nunes, Universidade de Coimbra, examinateur
  • Augustin Lux, INP Grenoble, examinateur
  • Christian Laugier, Inria Grenoble, directeur de thèse
  • Javier Ibanez-Guzman, Renault, co-encadrant

Intersections are the most complex and dangerous areas of the road network. Statistics show that most road intersection accidents are caused by driver error and that many of them could be avoided through the use of Advanced Driver Assistance Systems. In particular, vehicular communications open new opportunities for safety applications at road intersections. The sharing of information between vehicles over wireless links allows vehicles to perceive their environment beyond the field-of-view of their on-board sensors. Thanks to this enlarged representation of the environment in time and space, situation assessment is improved and dangerous situations can be detected earlier. This thesis tackles the problem of risk estimation at road intersections from a new perspective : a framework is proposed for reasoning about traffic situations and collision risk at a semantic level instead of at a trajectory level. Risk is assessed by estimating the intentions of drivers and looking for conflicts in them, rather than by predicting the future trajectories of the vehicles and looking for intersections between them. The proposed approach was validated in field trials using passenger vehicles equipped with vehicle-to-vehicle wireless communication modems, and in simulation. The results demonstrate that this algorithm allows the early detection of dangerous situations in a reliable manner and complies with real-time constraints. The proposed approach differs from previous works in two key aspects. Firstly, it does not rely on trajectory prediction to assess the risk of a situation. Dangerous situations are identified by comparing what drivers intend to do with what they are expected to do according to the traffic rules and the current context. The reasoning about intentions and expectations is performed in a probabilistic manner to take into account sensor uncertainties and interpretation ambiguities. Secondly, the proposed motion model includes information about the situational context. Both the layout of the intersection and the actions of other vehicles are taken into account as factors influencing the behavior of a vehicle.