Protection of Vulnerable Road Users

The main goal of this sub-project is development of user-friendly, cognitive systems in passenger cars and utility vehicles for effective and predictive protection of vulnerable road users in complex urban scenarios. In passenger cars, the focus is on active and forward-looking systems with a potential for timely detection of hazardous situations involving vulnerable road users (VRU) prior to an acute accident threat. These systems could then trigger appropriate protective measures for mitigating accident severity or avoiding a collision entirely. For utility vehicles, a system for substantial improvement of all-around visibility in hazardous situations is being developed. 


The research activities in the sub-project are concentrating on an inner-city scenario -- i.e., near-field and intermediate-range sensors, vehicle speeds up to 60 km/h or about 35 mph -- since a large proportion of accidents with VRU occur under these circumstances. Systems need to be capable of both daytime and nighttime detection. The requirements pose severe challenges to sensors and algorithms: the considerable variation in the visual appearance of humans in traffic (different shapes, levels of contrast); the highly dynamical and variable movements; the limited visibility due to frequent occlusion of persons by objects, ranging from baby carriages to parked vehicles; and the inherent complexity of many relevant accident scenarios.

Key technologies contributing to overall robust performance of such protective systems involve detection of road users by sensor-based systems such as video or radar and relevance testing.  Protective systems should respond appropriately in every situation; however, the appropriate response could be a warning in one situation or an autonomous action in another. The challenge of deciding among these options requires algorithms capable of classifying and characterizing the situation in depth, including object-specific behavior prediction.

The research project is mainly focused on improving the effectiveness and robustness of forward-looking systems in VRU protection, compared to the state of the art, by:

  • Addressing a much larger spectrum of use cases, i.e., VRU accident scenarios that were previously not detectable.
  • Improving system availability through a range of weather conditions;
  • Development of high-performance algorithms for sensor-based detection and classification, as well as modeling and prediction of behavior regarding vulnerable road users in all addressed scenarios and use cases.