Influence of the Pedestrian Geometry in Collision Avoidance Systems

The majority of fatalities between cars and pedestrians occur in urban areas. One approach to reduce these accidents are cooperative collision avoidance systems, in which cars and pedestrians exchange movement information to detect collisions. Typically, a scenario in such systems is modelled in two dimensions using a bird's eye perspective. One of the challenges in designing a collision avoidance system is the modelling of pedestrians for collision detection. Depending on which geometry is used for the pedestrian, it can influence the collision detection performance, leading to the risk that collisions are not detected correctly. This risk is further increased by sensor errors, resulting in inaccurate position and movement information. To address these challenges, we propose the following contributions in this work. First, we propose a novel set of collision scenarios based on the assessment protocol for Vulnerable Road User Protection of the European New Car Assessment Programme (Euro NCAP) and extend it by non-collision scenarios. Second, we introduce a new metric (detection score) which measures the overall collision detection performance based on our proposed set of scenarios. Third, we simulate and analyse how different shapes and sizes for the pedestrian model influence the detection score.