On the Required Movement Recognition Accuracy in Cooperative VRU Collision Avoidance Systems

Every year, approximately 350,000 vulnerable road users (VRU), like pedestrians and bicyclists, die due to collisions with vehicles. One approach for reducing this number of fatalities is using VRU collision avoidance systems, which aim at detecting and preventing possible collisions. Existing products for VRU collision avoidance systems using sensors mounted on vehicles may already decrease the number of accidents. However, these products require a line-of-sight between the vehicle and a VRU. Especially in urban environments, the line-of-sight between a vehicle and a VRU is often obstructed, e.g., by parked cars. Cooperative VRU collision avoidance systems promise to be a solution to this limitation. In such cooperative collision avoidance systems, the VRU is equipped with a portable device like a smartphone or a smartwatch. This allows exchanging the movement information between the vehicle and the VRU. However, one open question is how accurately such a system needs to recognize the movement of the VRU to achieve the necessary system performance. In this article, we provide a comprehensive analysis of how the accuracy of recognizing a pedestrian' s movement influences the reliability of detecting collisions between the car and the pedestrian in cooperative VRU collision avoidance systems. We conduct this analysis for the "Car-to-Pedestrian Nearside" scenarios from the "Test protocol for Autonomous Emergency Braking VRU systems" of the European New Car Assessment Programme (Euro NCAP).