Evaluation of a Collaborative-Based Filter Technique to Proactively Detect Pedestrians at Risk

Tragically, traffic accidents involving pedestrians or cyclists cause thousands of fatalities and serious injuries world- wide every year. Therefore, improving the safety of vulnerable road users is an international priority. One key challenge in designing an "ideal" protection system is to filter the endangered pedestrians out of potentially many. In this paper, we present a novel approach to proactively filter those pedestrians whose very next step would bring them (dangerously) closer to the street so as to provide an extra and crucial time advantage for a collision avoidance system. To predict a pedestrian\´s next step, we use the Collaborative Context Predictor. It takes advantage of collaborative behaviour patterns, in this case the movement patterns of the pedestrians. As well, a comparison with two state of the art context prediction approaches is carried out. The comparison is performed using real measured movement data, which is received from a smartphone the pedestrians carry in their trouser pocket.