In this paper an approach to predictive pedestrian protection is being proposed. The main issues regarding the identification of high benefit scenarios, the requirements for an appropriate risk assessment algorithm as well as the requirements for the environmental sensor system are discussed. A general survey of the topic is given first, including accident statistics regarding vulnerable road users. Based on more detailed accident data the requirements for a video-based pedestrian recognition system are derived. As a result the best suited aperture angle for early detection of pedestrians was determined.

A possible approach for predictive pedestrian protection is to issue an adequate driver warning in case of an impending vehicle-pedestrian collision. In order to justify driver warnings it is necessary to calculate the collision risk with a relatively large time-foresight. To cope with this task a pedestrian motion model based on likely and possible accelerations has been developed.