抄録
Estimating the benefit of advanced safety systems before launch is useful to develop and enhance the systems effectively. Several estimation methods based on accident databases have been proposed. However, these approaches have limitations because of the deficiency in either number of accident samples or kinetic details. We proposed a novel method that break through the limitations; generative method that enables us to acquire virtual rear-end accidents as many as necessary. The key is the finding through accident analysis; The majority of rear-end accidents are caused not by an exceptional driving behavior, but by a combination of normal driving behavior and a bit of inattention. Based on the idea, a normal traffic flow without any accident is firstly created. The ASSTREET (Advanced Safety System & Traffic REaltime Evaluation Tool) was used to simulate the road environment, vehicle movements and also driver behaviors. This traffic flow is considered to have a natural distribution of vehicle intervals, velocities and their combinations. Secondly, copy an arbitrary pair of vehicles that are in leading庸ollowing relation from the flow to another independent road environment. Then simulate what if the following car driver starts innatation at the timing. The safety systems' benefits can also be estimated comparing the results with and without system behavior. The advantage of the method is that a mass of rear-end collisions enables to evaluate system differences quantitatively and clearly points out the region that the system is expected to work effectively.
