It is commonly agreed that active safety will have a significant impact on reducing accident figures for pedestrians and probably also bicyclists. However, chances and limitations for active safety systems have only been derived based on accident data and the current state of the art, based on proprietary simulation models. The objective of this article is to investigate these chances and limitations by developing an open simulation model. This article introduces a simulation model, incorporating accident kinematics, driving dynamics, driver reaction times, pedestrian dynamics, performance parameters of different autonomous emergency braking (AEB) generations, as well as legal and logical limitations. The level of detail for available pedestrian accident data is limited. Relevant variables, especially timing of the pedestrian appearance and the pedestrian's moving speed, are estimated using assumptions. The model in this article uses the fact that a pedestrian and a vehicle in an accident must have been in the same spot at the same time and defines the impact position as a relevant accident parameter, which is usually available from accident data. The calculations done within the model identify the possible timing available for braking by an AEB system as well as the possible speed reduction for different accident scenarios as well as for different system configurations. The simulation model identifies the lateral impact position of the pedestrian as a significant parameter for system performance, and the system layout is designed to brake when the accident becomes unavoidable by the vehicle driver. Scenarios with a pedestrian running from behind an obstruction are the most demanding scenarios and will very likely never be avoidable for all vehicle speeds due to physical limits. Scenarios with an unobstructed person walking will very likely be treatable for a wide speed range for next generation AEB systems.
In spite of today's highly sophisticated crash test procedures like the different NCAP programs running world-wide, bad real world crash performance of cars is still an issue. There are crash situations which are not sufficiently represented by actual test configurations. This is especially true for car to car, as well as for car to object impacts. The paper describes reasons for this bad performance. The reasons are in principal bad structural interaction between the car and its impact partners (geometric incompatibility), unadjusted front end stiffness (stiffness incompatibility) and collapse of passenger compartments. To show the efficiency of improving cars' structural behaviour in accidents with different impact partners an accident data analysis has been taken out by members of European Project VC-COMPAT. Accident data analysis has shown that in Germany between 15,000 and 20,000 of the now severely injured car occupants might get less injured and between 600 and 900 car occupant fatalities might be saved. Similar results arise for the UK.